Fun for @AndrewLBeam and me to get a chance to chat with @suchisaria today for @NEJM_AI Grand Rounds!
Suchi Saria
Suchi Saria is an Associate Professor of Machine Learning and Healthcare at Johns Hopkins University, where she uses big data to improve patient outcomes. She is a World Economic Forum Young Global Leader. From 2022 to 2023, she was an investment partner at AIX Ventures. AIX Ventures is a venture capital fund that invests in artificial intelligence startups. More on Wikipedia
-
Suchi Saria reposted
Arjun (Raj) Manrai
@arjunmanrai
-
Suchi Saria reposted
JHU Computer Science
@JHUCompSci
In @HealthcareBrew: “Ultimately, we can only produce outcomes when clinicians trust the system and act on it,” says @bayesianhealth founder @suchisaria. “You can think of the @US_FDA as a very meaningful and critical level for building trust.” www.healthcare-brew.com/stories/bayesian-health-gets-fda-…
-
Suchi Saria
@suchisaria
Breakfast with some current and former students at ICML — always a great time!
-
Suchi Saria reposted
JHU Malone Center for Engineering in Healthcare
@JHUMCEH
An #ai system developed by John C. Malone Associate Professor of Computer Science @suchisaria detects deadly infections faster than doctors, saving thousands from sepsis—a condition that claims more than 250,000 lives each year in the U.S. malonecenter.jhu.edu/fda-approves-early-warning-system-fo…
-
Suchi Saria reposted
JHU Malone Center for Engineering in Healthcare
@JHUMCEH
“Once a clinician already suspects sepsis, the clock has been running—often for hours or even days,” says lead researcher and John C. Malone Associate Professor of Computer Science @suchisaria. “No other cleared test or device monitors for sepsis prior to clinician suspicion.”
Johns Hopkins University
@JohnsHopkins
The @US_FDA has approved an AI-powered early warning system developed by Johns Hopkins researchers that can detect sepsis before clinicians may even suspect it. hub.jhu.edu/2026/05/12/fda-approves-early-warning-system-… #researchsaveslives
-
Suchi Saria reposted
JHU Computer Science
@JHUCompSci
SPOT💡: In “Conformal Policy Control,” @DrewPrinster, @anqi_liu33, @suchisaria, @clara_fannjiang, @jiwoncpark, @kchonyc, & @samuel_stanton_ show how to use any safe reference policy as a probabilistic regulator 4 any optimized but untested policy: arxiv.org/abs/2603.02196 (14/25)
-
Suchi Saria reposted
Drew Prinster
@DrewPrinster
WATCH: Adaptive Monitoring for AI Deployments via Weighted-Conformal Martingales (arxiv.org/abs/2505.04608), w/ @xinghan0, @anqi_liu33, @suchisaria:
- Sat (7/11, Room 318) ICML Workshop on Hypothesis Testing
(3/6) -
Suchi Saria reposted
JHU Computer Science
@JHUCompSci
“Improving Coverage in Combined Prediction Sets with Weighted p-values” by @DrewPrinster, @suchisaria, @ChellappaProf, @anqi_liu33, & more proposes a framework for the weighted aggregation of prediction sets w/ weights assigned based on contribution: arxiv.org/abs/2505.11785 (6/7)
-
Suchi Saria reposted
Drew Prinster
@DrewPrinster
Can we ensure AI agents respect our safety constraints, even as they explore & improve?
- Medical LLMs that are helpful, & avoid false claims?
- Bioscience agents that generate effective molecule designs, & ensure they’re safe?
📄🧵w/ @samuel_stanton_ @clara_fannjiang @jiwoncpark @kchonyc @anqi_liu33 @suchisaria
Excited to share “Conformal Policy Control” ⬇️
1/12 -
Suchi Saria reposted
Chris Bakke
@ChrisJBakke
Running a company:
2020: can you survive a pandemic?
2021: still here? we’re going to give all of your competitors $100m series A rounds.
2022: wow, you made it? okay, all engineers cost $600,000/year now.
2023: nice job! okay, SVB failed and we’re going to take away your bank account.
2024: a survivor I see. but can you pivot from ai to crypto to defense tech back to ai-enabled defense tech in a 12 month period to stay relevant?
2025: unfortunately all of your competitors have raised $2b series B rounds. oh and only 500 engineers are relevant and they cost $100m/yr each.
2026: well, well, well. you’re still in business? let’s deploy the thunderclap of godlike LLMs from the heavens so all of your customers can rebuild your app in 2 hours. can you survive? -
Suchi Saria reposted
JHU Computer Science
@JHUCompSci
As #ai transforms nearly every sector, @suchisaria outlines guiding principles in @NatureMedicine for AI’s use in medical care so that it can truly benefit physicians and patients.
Tej D. Azad 🇺🇸
@tdazad
Clinical AI doesn’t need better AUC. It needs better evaluation design.
Benchmarks ≠ bedside readiness. Better evaluation will build clinician trust.
New perspective in @NatureMedicine w/ @suchisaria & @hmkyale
Link: rdcu.be/e0IkI -
Suchi Saria
@suchisaria
I’m in Chicago for work and realizing I have extra time open this evening! What should I do? Curious to hear ideas.
-
Suchi Saria reposted
Doctor Radio
@NYUDocs
Could #ai help predict #sepsis? At 2:30pm join @DrMarcSiegel as he speaks with @suchisaria (AI prof/Endowed Chair @JohnsHopkins & Founder @BayesianHealth) and @withyouDrWu (Professor, director of CHSOR @JohnsHopkinsSPH) about their research. Stream here: sxm.app.link/DoctorRadio
-
Suchi Saria reposted
Drew Prinster
@DrewPrinster
AI monitoring is key to responsible deployment. Our #icml2025 paper develops approaches for 3 main goals:
1) *Adapting* to mild data shifts
2) *Quickly Detecting* harmful shifts
3) *Diagnosing* cause of degradation
🧵w/ @xinghan0 @anqi_liu33 @suchisaria
arxiv.org/abs/2505.04608 -
Suchi Saria
@suchisaria
🚨With an awesome group of collaborators from health system & federal agencies, @Jean_J_Feng and I are embarking on an ambitious multi-year project to develop tools for monitoring & updating clinical AI algorithms, with the aim of informing how we accelerate & scale responsible AI deployment. More details here jeanfeng.com/joining Looking postdocs, data scientists and senior research fellow applications. Apply!
Thanks for RT. -
Suchi Saria reposted
JHU Computer Science
@JHUCompSci
Awesome highlights of the important research being done by @MathiasUnberath and @suchisaria! #researchsaveslives
Johns Hopkins University
@JohnsHopkins
From diagnosing infections with a smartphone to training surgical robots, Johns Hopkins researchers are using AI to create real-world impact. hub.jhu.edu/research-saves-lives/
-
Suchi Saria
@suchisaria
“For rare disease, the journey is 7 years before diagnosis” - Dr. Janet Woodcock
wow.🤯 -
Suchi Saria reposted
Johns Hopkins Engineering
@HopkinsEngineer
Hopkins Engineering faculty Rama Chellappa, @suchisaria, and Axel Krieger talk AI healthcare with @ScienceNews!
Science News
@ScienceNews
AI could transform health care, but the stakes are high. www.sciencenews.org/article/ai-health-care-medicine-tech
-
Suchi Saria reposted
Drew Prinster
@DrewPrinster
When do AI explanations actually help, & promote appropriate trust?
Spoiler, via prospective, multisite @radiology_rsna study of 220 doctors: *How* AI explains its advice has big impacts on doctors’ diagnostic performance and trust in AI--even if they *don’t realize it*!
🧵1/ -
Suchi Saria reposted
Drew Prinster
@DrewPrinster
Paper 🧵! w/ @samuel_stanton_ @anqi_liu33 @suchisaria #icml2024 arxiv.org/abs/2405.06627
We study 2 Qs:
1) Can AI uncertainty quant. via #conformalprediction extend to any data distribution?
2) Are there practical CP algorithms for AI/ML agents 🤖w/ feedback-loop data shifts?
1/ -
Suchi Saria
@suchisaria
Thank you 🙏 cc @DrewPrinster
Valeriy M., PhD, MBA, CQF
@predict_addict
This is a monumental breakthrough in the realm of Conformal Prediction, akin to discovering another universe in physics! 🔥🔥🔥🔥🔥
A groundbreaking new paper from John Hopkins University and Genentech titled "Conformal Validity Guarantees Exist for Any Data Distribution" by Drew Pinster, Samuel Stanton, Anqi (Angie) Liu, Suchi Saria has proven that conformal prediction guarantees are applicable to ANY data distribution, not just exchangeable ones. 🚀🚀🚀🚀🚀
This fundamental advancement paves the way for entirely new areas of research and practical applications.
#conformalprediction -
Suchi Saria reposted
Drew Prinster
@DrewPrinster
Had a (positive) “deer-in-the-headlights” moment today when scrolled to these kind & generous words from @predict_addict on our new paper w/ @samuel_stanton_ @anqi_liu33 @suchisaria! 😆arxiv.org/abs/2405.06627
Humbled & grateful to share this work. Full paper summary thread soon!
Valeriy M., PhD, MBA, CQF
@predict_addict
This is a monumental breakthrough in the realm of Conformal Prediction, akin to discovering another universe in physics! 🔥🔥🔥🔥🔥
A groundbreaking new paper from John Hopkins University and Genentech titled "Conformal Validity Guarantees Exist for Any Data Distribution" by Drew Pinster, Samuel Stanton, Anqi (Angie) Liu, Suchi Saria has proven that conformal prediction guarantees are applicable to ANY data distribution, not just exchangeable ones. 🚀🚀🚀🚀🚀
This fundamental advancement paves the way for entirely new areas of research and practical applications.
#conformalprediction -
Suchi Saria
@suchisaria
‘Nuf said.
"Leigh Marie" Braswell
@LM_Braswell
Broke out some S tier @scale_AI swag (from the vault)
-
Suchi Saria reposted
Johns Hopkins University
@JohnsHopkins
Johns Hopkins and Columbia University computer scientists have proposed a method to enhance the robustness of AI models used in medical text analysis. The project is part of an ongoing effort to develop an AI safety framework for healthcare applications. hub.jhu.edu/2024/02/20/artificial-intelligence-analyzes-d…
-
Suchi Saria reposted
JHU Computer Science
@JHUCompSci
Computer scientists from @JohnsHopkins @HopkinsEngineer (featuring @suchisaria) and @Columbia combat the inaccurate correlations that #ai learns from text data. Learn more: hub.jhu.edu/2024/02/20/artificial-intelligence-analyzes-d…
-
Suchi Saria
@suchisaria
Excited for us at @HopkinsEngineer to be part of this!
Dr. Darryll J. Pines
@President_Pines
A great pleasure partnering with @CommerceGov @SecRaimondo and over 200 stakeholders to launch the U.S. Artificial Intelligence Safety Consortium with @NIST. The largest initiative of its kind, #aisic will advance safe, secure, and trustworthy AI.
-
Suchi Saria
@suchisaria
#1stworldproblems #1stworldjokes
After years, I recently got downgraded from United Global Services to 1K.
The air hostesses think they’re being nice when they greet me — “You’re a 1K member! Thank you for your loyalty.” It mostly hurts, feels like they’re rubbing it in! 😂😂 -
Suchi Saria
@suchisaria
Will your LLM model generalize over time or across sites? How do we improve robustness? This is a key issue LLMs struggle with and especially critical as we apply these models to healthcare applications.
👇 New clever technique to improve LLM robustness and generalization. Published @NeurIPSConf w/ wonderful postdoctoral fellow @wald_yoav & superb co-authors from Columbia.
engineering.jhu.edu/news/in-others-words-using-large-lang…
arxiv.org/abs/2310.12803 -
Suchi Saria
@suchisaria
This is what rich people struggling looks like: wine parties w/o wine & billionaires sleeping on couches ;) www.wsj.com/lifestyle/travel/bad-hotels-and-not-enough-wi…
And apparently everybody is leading the global conversation on #ai! 😂 😂 😂 -
Suchi Saria reposted
Avani K. Patel, MD, MHA
@AvaniPatelMD
Absolutely impressive panel ( @AnthonyChangMD @suchisaria PhD, Mark Sendak, MD, & Patrick Tighe, MD) at @AmerMedicalAssn #stateadvocacysummit who share about AI: the promise; the potential, the needed guardrails
-
Suchi Saria reposted
Marilyn Heine
@MarilynHeineMD
⭐️#stateadvocacysummit
🌠Incredible lineup of guest keynote and panel speakers to address hot topics🔥
🌠Exciting new state advocacy campaign and an AMA advocacy update.
Looking forward to amazing conversations!
More info: 👉 www.ama-assn.org/membership/events/state-advocacy-summit
#ourama #fightingfordocs -
Suchi Saria reposted
a16z Bio + Health
@a16zBioHealth
Yesterday at @jpmorgan's Healthcare Conference, a16z Bio+Health GP @julesyoo dove into what the year ahead will bring for AI in healthcare alongside industry leaders @suchisaria of Bayesian Health, @sethHain of Epic, and @peteratmsr of Microsoft. #jpm24
ICYMI, check out some of Julie’s thoughts on the future of AI in healthcare with her recent publications:
👉🏼 Accessing, Paying for, and Delivering Cures, at Scale: a16z.com/the-new-drug-delivery-challenge-how-will-cures-r…
👉🏼 The Next WebMD: An LLM as Your Front Door to Healthcare: a16z.com/the-next-webmd-an-llm-as-your-front-door-to-heal…
👉🏼 Commercializing AI in Healthcare: The Enterprise Buyer Perspective:
a16z.com/commercializing-ai-in-healthcare-the-enterprise-…
Julie Yoo
@julesyoo
I'll be chatting with @suchisaria of @bayesianhealth, @sethHain of Epic, and @peteratmsr of @Microsoft about "AI in Healthcare: The Opportunity Ahead" - Monday 5:15 pm, in the Colonial Room at #jpm2024.
We'll cover the state of the state, regulatory landscape, and predictions and advice for healthcare execs re: healthcare AI as we head into the new year ahead.
@a16zBioHealth -
Suchi Saria reposted
Yoav Wald
@wald_yoav
Drop by our #neurips2023 poster (#910) today (Wed.) at 5PM!
We improve OOD generalization in text applications with Language Models + Causal-inference-inspired data augmentation.
w/ best possible co-lead @AmirFeder , and @causalclaudia, @suchisaria, David Blei -
Suchi Saria reposted
Vijay Pande
@vijaypande
One of the key highlights is that @suchisaria gives her take on the implications for healthcare of the recent Exec Order on AI.
Very timely and thoughtful take on how we can work together to improve healthcare with AI.
a16z.com/podcast/clinical-copilots-augmenting-human-care-…
a16z Bio + Health
@a16zBioHealth
In the era where AI meets Health, navigating the regulatory framework is not just a compliance chore, but a gateway to trust and broader adoption. @suchisaria, Founder & CEO of @bayesianhealth, and @vijaypande, general partner of a16z Bio+Health, break it down:
👉🏼 Embrace oversight! It's not a hurdle but a facilitator that ensures the safety and efficacy of AI applications, accelerating their adoption in healthcare.
👉🏼 The future is real-time. Building an infrastructure that collects and computes relevant data in real-time paves the way for actionable and intelligent healthcare solutions, transforming care delivery.
👉🏼 Beyond deployment, the journey continues. Continuous monitoring and real-time tuning of AI tools are pivotal to ensure they remain effective and safe amidst the ever-evolving healthcare landscape.
Listen to the full episode on Bio Eats World: a16z.com/podcast/clinical-copilots-augmenting-human-care-… -
Suchi Saria
@suchisaria
📣 📣 2 new full stack data scientist and AI/ML engineer positions newly opened @bayesianhealth!
We won TIME Best Inventions, Forbes AI50, World Economic Forum Tech Pioneer.
Amazing team + great purpose + fun/hard AI problems.
Apply!
www.bayesianhealth.com/careers/
Suchi Saria
@suchisaria
🎉 🎉 🎉 @bayesianhealth selected to TIME Best Inventions 2023!
Not going to lie… getting selected is awesome but what I love even more is how they characterize the impact.
If you’re a backend / data / AI/ML engineer, and want to build AI copilots for doctors and nurses to improve care, apply. It’s an incredible mission, solid team, and a culture I’m proud of!
time.com/collection/best-inventions-2023/6324389/targeted… -
Suchi Saria reposted
JAMA
@JAMA_current
Suchi Saria, PhD, MSc, computer scientist at @JohnsHopkins, discusses the use of AI tools in early sepsis detection and other clinical applications.
Watch “AI and Clinical Practice" with @suchisaria and @KBibbinsDomingo.
#jamaai
jamanetwork.com/channels/ai?linkId=245262297 -
Suchi Saria reposted
Johns Hopkins Engineering
@HopkinsEngineer
.@suchisaria joined @KBibbinsDomingo on @JAMANetwork's “AI and Clinical Practice” series, discussing how machine learning and big data can individualize patient care and save lives. Her AI-based sepsis detection tool is a TIME Best Invention of '23. Watch: jamanetwork.com/journals/jama/fullarticle/2811547
-
Suchi Saria reposted
Kirsten Bibbins-Domingo
@KBibbinsDomingo
“..going from tools in the lab setting where you’re taking on retrospective data & showing what’s possible to implementing it in a practical clinical setting…to influence care in a positive way is where a lot of new research needs to happen” @suchisaria
jamanetwork.com/journals/jama/fullarticle/2811547 -
Suchi Saria
@suchisaria
Tomorrow @POTUS President Biden’s executive order on AI is set to be released. I got to be a part of some of the meetings leading up to it and had a chance to read a document dubbed as early draft.
Thrilled to see the administration focusing on this important area.
For a number of sectors including healthcare, a lot of the follow up activity — eg aligning on and putting best practice rubrics in place — will occur over the next 180 and 365 days! Through public-private partnerships like the @theNAMedicine National Academy of Medicine’s AI Code of Conduct (I’m on the taskforce) and CHAI, the Coalition of Health AI (COI: I’m one of the cofounders), work of this type has been underway already for 1+ year.
The EO (at least in its draft version) doesn’t spell out a point of view on some critical matters like who gets to call out whether an org is following guardrails? How do we guard against stifling innovation? Example adverse possibilities would be a significant increase in cost of development, delays because of inefficient approval processes, or having mega vendors/incumbents become gate keepers.
Looking forward to reading reactions to the executive order.
www.politico.com/news/2023/10/27/white-house-ai-executive… -
Suchi Saria
@suchisaria
🎉 🎉 🎉 @bayesianhealth selected to TIME Best Inventions 2023!
Not going to lie… getting selected is awesome but what I love even more is how they characterize the impact.
If you’re a backend / data / AI/ML engineer, and want to build AI copilots for doctors and nurses to improve care, apply. It’s an incredible mission, solid team, and a culture I’m proud of!
time.com/collection/best-inventions-2023/6324389/targeted… -
Suchi Saria reposted
Danielle Carnival Archived
@DCarnival46
As @POTUS has made clear, we are committed to realizing the benefits of AI for all Americans but need to understand & mitigate its risks. As a shared responsibility, it is essential to bring together leaders to ensure AI is used to equitably improve health care & health outcomes.
White House Office of Science & Technology Policy
@WHOSTP
This week, a White House roundtable on AI and health brought together leaders across industries for a robust discussion about how AI can deliver benefits in patient care, drug development, and public health. Read more: www.whitehouse.gov/ostp/news-updates/2023/10/06/readout-o…
-
Suchi Saria reposted
Arundhati Parmar
@aparmarbb
#hlth2023 is around the corner & @medcitynews is ready with ENGAGE at HLTH patient engagement track on Sunday, Oct. 8 from 10 a.m.- 1 p.m. Hear from @AmazonHealthSVC's Sunita Mishra @generalcatalyst's @AlexandreMomeni and @Bayesianhealth's @suchisaria. Moderateor@ravindranize
-
Suchi Saria reposted
Brian Wells
@bwells80
#rt @ctisus: RT @TinglongDai: 🌟Thrilled to join Dean Theodore DeWeese of @HopkinsMedicine and esteemed colleagues from across @JohnsHopkins for today's AI in Healthcare panel. With its legacy as a pioneer in modern medicine, @JohnsHopkins is leading …
-
Suchi Saria
@suchisaria
If you’re at @IJCAIconf please join me for my invited keynote Thursday 2pm Hong Kong time. I’ll be speaking about AI for human-machine teaming in high stakes applications like medicine and what’s necessary to safely and responsibly implement health AI: ijcai-23.org
DM if you’d like to grab dumplings!Post not available here, view on X
Suchi Saria
@suchisaria
One of her new focus areas is to build automated mediators. Why? Because “people are increasingly arguing all the time” and the hope is these mediators can help. Could be very interesting!
Suchi Saria
@suchisaria
Next talk is by this year’s IJCAI Computer and Thought award winner Pin-Yu Chen @pinyuchenTW
Excited to hear about his work. -
Suchi Saria reposted
Yoav Wald
@wald_yoav
Our #uai2023 paper is on detecting novel subgroups (aka classes/categories) when distribution of non-novel data shifts.
If this sounds interesting, or you’re wondering what it has to do with unicorns, check out our poster/paper/drop a line.
w/ @suchisaria
proceedings.mlr.press/v216/wald23a.html -
Suchi Saria
@suchisaria
If you’re at #icml2023 join me here: sites.google.com/view/scis-workshop-23/home?authuser=0
Giving an invited talk later today on AI Safety.
Below are some of the papers our lab has written on the topic in last 5 yrs; this talk will give a brief overview and focus on learning/evaluation.
Suchi Saria
@suchisaria
Some overview articles that I missed adding:
www.nejm.org/doi/full/10.1056/NEJMc2104626
academic.oup.com/biostatistics/article/21/2/345/5631850
arxiv.org/pdf/1904.07204.pdf -
Suchi Saria reposted
Drew Prinster
@DrewPrinster
Our Oral presentation is today (Wed 7/26) 4-5:30 p.m. HST in meeting Room 316 A-C for anyone at #icml #icml2023 and interested! :)
(We're in the Oral B3 Privacy Session, though our paper is on UQ/#conformalprediction 🙃🤷♂️. Still grateful for the opportunity to share our work!)
Drew Prinster
@DrewPrinster
Can I trust this AI/ML prediction? A good uncertainty quantification (UQ) method could answer this with a predictive confidence interval. But in practice, 2 key challenges limit real-world UQ.
Our #icml2023 Oral “JAWS-X” tackles these challenges! 🧵w/ @suchisaria @anqi_liu33 1/n -
Suchi Saria reposted
Drew Prinster
@DrewPrinster
Can I trust this AI/ML prediction? A good uncertainty quantification (UQ) method could answer this with a predictive confidence interval. But in practice, 2 key challenges limit real-world UQ.
Our #icml2023 Oral “JAWS-X” tackles these challenges! 🧵w/ @suchisaria @anqi_liu33 1/n -
Suchi Saria
@suchisaria
This is such an interesting example of — what I think of as — a great use case.
ABC News
@ABC
One man who has communicated with the digital reproduction of his mom on a daily basis since she died in October, believes AI-driven chatbots will redefine how some deal with grief. abcnews.com/Business/ai-advances-fuel-industry-preserve-l…
-
Suchi Saria reposted
a16z
@a16z
1 in 3 nurses is leaving the workforce. AI can dramatically streamline healthcare workflow, meaning less admin work & better patient outcomes.
We talked to top AI founders, including @suchisaria, about what they're building.
Hear more in their words: a16z.com/ai -
Suchi Saria
@suchisaria
There’s been a lot of discussion around the need for guideline and rubrics for safe and responsible adoption of #ai in health. Today, the National Academy of Medicine launched this multi-year project and taskforce to address this challenge.
“The Artificial Intelligence Code of Conduct (AICC) project is a pivotal initiative of the National Academy of Medicine (NAM), aimed at providing a guiding framework to ensure that AI algorithms and their application in health, medical care, and health research perform accurately, safely, reliably, and ethically in the service of better health for all. Stewarded by the NAM Leadership Consortium, the project will yield a pioneering AI Code of Conduct framework reflecting best practices to serve as a starting point of reference for follow-on testing, validation, monitoring, and continuous improvement.”
In a sector like healthcare where mistakes can be costly, best practices are very helpful for accelerating adoption. It allows users to understand risk and invest in implementing new ideas more comfortably. It also gives federal agencies a blueprint to point to in guiding their own efforts. From the perch of the NAM, there is also an opportunity to align similar rubric development efforts across many grassroots and federal groups.
You can subscribe in the link here to follow along! nam.edu/programs/value-science-driven-health-care/health-… -
Suchi Saria
@suchisaria
So proud!
Bayesian Health
@bayesianhealth
🎉 🎉 🎉 We are thrilled to be selected as a World Economic Forum Technology Technology Pioneer!
This recognition brings together 100 startups from around the world that are "at the forefront of innovation and disruption needed to help us solve the world’s most pressing issues.” -
Suchi Saria reposted
Bayesian Health
@bayesianhealth
🎉 🎉 🎉 We are thrilled to be selected as a World Economic Forum Technology Technology Pioneer!
This recognition brings together 100 startups from around the world that are "at the forefront of innovation and disruption needed to help us solve the world’s most pressing issues.” -
Suchi Saria
@suchisaria
😢 Sadly this is commonplace in CS/AI/ML.
People tend to cite folks within their circle (friends/school). Given the number of wonderful female scholars studying AI ethics, this is a glaring example.
@emilymbender.bsky.social
@emilymbender
A bingo for Geoffrey Hinton's keynote at #acl2023nlp. Here are all the non-men he cited.
-
Suchi Saria reposted
Carlos del Rio
@CarlosdelRio7
Great fireside chat at @theNAMedicine on AI in biomedical research, healthcare delivery and health policy with @zakkohane @peteratmsr & @suchisaria
-
Suchi Saria
@suchisaria
❤️ this thread/letter👇.
I’ve read 2k+ faculty letters (fellowship apps, faculty apps). This letter is so special, not only because @mmitchell_ai is special but because of the care @timnitGebru takes in writing it.
She considers the *whole* story behind Meg’s work including calling out when a piece of work is phenomenal but didn’t get the recognition it deserves: “Her model was tied with that of Oriol Vinyals but if one reads the press from 2015, her name is nowhere to be found.”
Experts can/should leverage their perch to put context around a piece of work like Timnit does here. It’s tempting to go to g scholar and just pull out citations stats but these perpetuate existing biases in the way works get cited.
I also love how Timnit demands more of what it means to do high impact work:
“Many times, as researchers, we write these goals in the introduction section of our paper. However, Meg is one of the few people who worked with visually impaired people to ensure that her work was helpful to that community.” Most faculty don’t demand this of their own work and so as letter writers, they rarely show appreciation for the work it takes to *really* think sincerely about real-world impact.
Now imagine if the CS research community were more diverse (in just about every way) and we had 10% more scholars who brought Timnit’s perspective, I think we’d make much much faster progress!
@timnitGebru (@dair-community.social/bsky.social)
@timnitGebru
4-5 years ago @mmitchell_ai was a semifinalist for the MIT Tech Review 35 under 35. I wrote a letter of support. One of the things I mentioned was that her work has been so under appreciated in the field of "AI." 1/n
-
Suchi Saria
@suchisaria
My slides from #chil2023 YIR 2020-2022.
Covers some of my* favorite papers in #ai #ml & health and focus is on papers where I could clearly articulate why it is significant: www.dropbox.com/s/4ovbukk6dmhtnbe/CHIL - Saria - 2020-202…
Either,
- they helped significantly advance a key area, or
- made a key new idea known, or
- raised awareness on an important issue, or
- they show early work in an area where follow up is critically needed.
This deck isn’t to advertise any specific paper but more to give a sense of themes and high value areas. I had a chance to read through many good papers b/c of this talk. I’m sure I’ve missed others. Feel free to tag suggestions not covered in the talk to this thread.
*I had a wonderful committee help seed the shortlist; included several papers I’ve been part of. To be inclusive, we made multiple calls for nominations via twitter and other networks. Ultimately, as was the mandate by the OC, this talk heavily represents my POV. Enjoy!!! Hopefully, you’re able to take away at least one or two good ideas.
Suchi Saria
@suchisaria
On my way to #chil2023
Giving a talk tomorrow so will likely be a bit busy w/ prep today but I hope I’ll get to see many in person there!
@RoxanaDaneshjou @DrLaurenOR @oziadias @2plus2make5 @David_Ouyang @drnigam @GlockerBen @kdpsinghlab at the social this evening?
Who else is joining? -
Suchi Saria reposted
Dimitri Dadiomov
@dadiomov
“And since this is the last speech that I will give as President, I think it's fitting to leave one final thought, an observation about a country which I love. It was stated best in a letter I received not long ago. A man wrote me and said: ``You can go to live in France, but you cannot become a Frenchman. You can go to live in Germany or Turkey or Japan, but you cannot become a German, a Turk, or a Japanese. But anyone, from any corner of the Earth, can come to live in America and become an American.'”
This, I believe, is one of the most important sources of America's greatness. We lead the world because, unique among nations, we draw our people -- our strength -- from every country and every corner of the world. And by doing so we continuously renew and enrich our nation. While other countries cling to the stale past, here in America we breathe life into dreams. We create the future, and the world follows us into tomorrow. Thanks to each wave of new arrivals to this land of opportunity, we're a nation forever young, forever bursting with energy and new ideas, and always on the cutting edge, always leading the world to the next frontier. This quality is vital to our future as a nation. If we ever closed the door to new Americans, our leadership in the world would soon be lost.”
- Ronald Reagan, Jan 19, 1989
Happy Fourth! 🇺🇸 -
Suchi Saria
@suchisaria
At the JP Morgan Founders Forum. Loved sharing a main stage panel with a friend from 20 years ago, also the CEO of a kicka** high social impact AI company!
-
Suchi Saria
@suchisaria
Such an excellent and inspiring discussion 👇
clem 🤗
@ClementDelangue
This is my 5-minute testimony before the US Congress!
Open science and open source AI distribute economic gains by enabling hundreds of thousands of small companies and startups to build with AI. It fosters innovation, and fair competition between all.
Thanks to ethical openness, it creates a safer path for development of artificial intelligence by giving civil society, non-profits, academia, and policy makers the capabilities they need to counterbalance the power of big private companies.
Open science and open source AI prevent blackbox systems, make companies more accountable, and help solving today’s challenges like mitigating biases, reducing misinformation, promoting copyright, & rewarding all stake-holders including artists & content creators in the value creation process.
Let's go!This media couldn't be displayed. See it on the original post. -
Suchi Saria reposted
Matthew McDermott
@MattBMcDermott
Fantastic talks about defining papers and real world deployments in #ml4h at #chil23 this morning by @suchisaria and @kdpsinghlab! Lots of great ideas and insights to explore
-
Suchi Saria
@suchisaria
On my way to #chil2023
Giving a talk tomorrow so will likely be a bit busy w/ prep today but I hope I’ll get to see many in person there!
@RoxanaDaneshjou @DrLaurenOR @oziadias @2plus2make5 @David_Ouyang @drnigam @GlockerBen @kdpsinghlab at the social this evening?
Who else is joining?
Suchi Saria
@suchisaria
My goal at CHIL is to find 3 fun collaborators. The areas I’m excited about and very actively working in:
- AI safety and performance monitoring,
- Human-machine teaming/collaboration (eg explanations, predictive inference)
- Reliable LLM inference from messy observational data
- Real world evaluations
If you’re interested in these and want to brainstorm, come say hi! Of course, I’m happy to chat about anything fun! -
Suchi Saria reposted
Mustafa Suleyman
@mustafasuleyman
This is big news on EU AI regulation:
- ban on AI for biometric surveillance, emotion recognition, predictive policing
- registration of models with EU
- detailed summary of training data
- requirement to identify deepfakes
www.europarl.europa.eu/news/en/press-room/20230609IPR9621… -
Suchi Saria
@suchisaria
I’m putting together a year in review talk on AI/ML/informatics & health to be presented next Friday. It covers papers from 2020-2022. I’d LOVE to hear your favorite papers.
Focus is on: methods, real-world impact, high-impact promising applications.
Need to finalize by Wednesday! Pls RT. Thanks! -
Suchi Saria
@suchisaria
I’m a little confused by this ask. Whether or not a model is biased isn’t dependent on the input variables in a straight forward way — two models with the same inputs but different modeling approaches can yield models w/ more or little bias. Thoughts?
If we want to assess for bias, shouldn’t we explicitly measure if the model is biased? Also, ensure that groups deploying models are assessing for bias? Eg academic.oup.com/jamia/article/29/8/1323/6586579
The chance that third parties will spend the time to evaluate and identify bias based on input data info below is unlikely.
Asking for more also isn’t free. Perhaps we should identify how, with minimal extra reporting lift, we can achieve what’s maximally efficient for evaluating efficacy and bias.
Karandeep Singh
@kdpsinghlab
On April 18, @ONC_HealthIT issued a proposed rule w/ the goal of ⬆️ transparency on the use of healthcare algorithms.
@jeplatt, I, colleagues @UMich, & around the world support ⬆️ transparency.
We have 4 recs on being clear with *the public*.
Letter: downloads.regulations.gov/HHS-ONC-2023-0007-0024/attachme…
Suchi Saria
@suchisaria
Our overall current approach to improving oversight reminds me of this beautiful quote I heard recently:
Step 1 … we must do something
Step 2 … this is something
Step 3 … let’s do this
What if we studied what should a minimal reporting set look like that gives quality guarantees while ensuring minimal lift and preserving IP?
Suchi Saria
@suchisaria
I’m a huge fan of @kdpsinghlab work; intrigued to see you make these recs so curious to debate / discuss.
-
Suchi Saria reposted
Sanyam Bhutani
@bhutanisanyam1
Run 13B model on an iPhone! 🤯
Just finished reading @Tim_Dettmers’ amazing work on SpQR.
SpQR unlocks 3.35 bit quantisation which lets us run 33B models on 3090s and 13B models on an iPhone.
Here are my notes from the paper:
- Quantisation is basically like compressing the LLM weights
- These are normally trained at 16 or 8 bits and show performance loss at 4 bits
- An interesting insight is LLM parameters are not equally sensitive to quantisation
- Only 1% of the weights contribute to majority of the loss
- So if we can store these in 16-bits and compress everything else. We have lossless compression
- Following the above we can achieve lossless 4.75 bit and performant 3.35 bit performance.
This further gives a 4x size reduction of models and 20% speedup in inference. The best part-this happens post training.
Tim has shared both incredible work for during and post training quantisation. I’ll link his discussion in 🧵
Here’s the paper which goes into a detailed discussion of the algorithm and approach:
arxiv.org/abs/2306.03078 -
Suchi Saria
@suchisaria
If you’re a solid engineer w/ experience productionizing data products and you wanted to join grad school in AI/CS but didn’t get in, apply to work with @bayesianhealth
* You’ll work on hard & important multimodal time series problems,
* Leverage state of the art ideas and be coached to advance them,
* You’ll have a cross-disciplinary team of experts to learn from and collaborate with
* You’ll have access to large data and infra
* You’d have me as your mentor
I am confident you’ll learn as much and likely much more about AI as you might have learned in most labs.
** some candidates ask if we only use Bayesian techniques — no. The smartest decision makers integrate many types of data, think about how much to trust what, update their forecast in a principled way as new data arrives, and understand uncertainty; the company is named after this philosophy of reasoning. We have deep & broad AI expertise and as a result, use/design best in class techniques for the problems at hand.
Rosanne Liu
@savvyRL
I've been hearing from people that it's a very challenging year for AI/CS grad school application. A few words:
First, I'm sorry many of you have to live through this; it's never fun receiving rejections.
Second, I don't expect the situation to improve in foreseeable future.
Suchi Saria
@suchisaria
My lab’s PhD candidate acceptance rate is < 1% and we’ve trained several amazing graduates who are now professors at top schools or doing incredible work in industry.
In your work at Bayesian, you’d effectively be doing similar guided solution development. The big difference is that PhD work is heavily focused on publishing vs your work at the company will be heavily focused on real-world results. -
Suchi Saria
@suchisaria
I am getting at least 10-15 spam calls a day these days. There’s been a massive uptick in the last few months.
Is anyone else experiencing it too? If so, have you found a solution. -
Suchi Saria
@suchisaria
Excellent interview by @sharongoldman and @kchonyc
Especially love the point — contrary to how the media & outsiders tend to glorify a small number of individuals, AI progress has resulted from the collective efforts of many.
This inclusive “wisdom of the crowds” culture also ensures rapid progress, like we have seen in the post chatGPT release era.
Sharon Goldman
@sharongoldman
More AI "doomer" pushback. Thanks to @kchonyc for talking to me for this one. venturebeat.com/ai/top-ai-researcher-dismisses-ai-extinct…
Suchi Saria
@suchisaria
Re: regulation, the FDA’s made so much progress. (COI: I’ve been doing funded research with them for the past 3 years.)
The ideas and frameworks for regulation in health are so much more mature than ideas surfaced in that senate hearing. -
Suchi Saria reposted
Massimo
@Rainmaker1973
Jacaranda blooming season is one of the most beautiful times of year. This is what Johannesburg, South Africa looks like when that happens
[read more: en.wikipedia.org/wiki/Jacaranda] -
Suchi Saria reposted
Shahin Farshchi
@Farshchi
1/ Buzzy AI companies are raising giant dollars to train giant models
Giant models will indeed power magical demos and be useful to lots of lay consumers
... while the smaller, specialized models will be where the AI rubber hits the road. Here’s why 🧵 -
Suchi Saria
@suchisaria
Are you working in explainable ML/AI for hypothesis generation in life sciences? Ideally, you’re working within biotech, pharma, a startup, or a lab collaborating around life science discovery tasks. I have an ask for a NeurIPS talk but I’d love to invite others instead. Thanks!
-
Suchi Saria
@suchisaria
There is an AI discussion ask — whether it’s from the gov, VCs, or National collaboratives — 5-7 times a week!
Reminds me of the last AI wave… for a while, you had to make a decision about whether to do the work or inform the conversation.
I like to, and I imagine many researchers, are prioritizing research/doing-the-work and so we are left with conversations often led by those less informed.
There are so many amazing researchers & practitioners that I’d love if they were more involved! I’ve been forwarding asks I get … but need more leaders who care about the field/doing-the-right-thing over personal agenda (directing money to themselves / their institution / or becoming individually well-known). Do I sound like a cynic?
Suchi Saria
@suchisaria
Just got this one. Any experts (practitioners or researchers) who are interested?
-
Suchi Saria
@suchisaria
Headed to our 3rd full team retreat at @bayesianhealth 🌵🏜️💃
There will be rap battles, iron chef, dance competition, pool games, an award night and loads of team bonding over 4 days.
I remember 3 years ago struggling to understand how to build a company with a great culture. I feel so proud of where we’ve come.
This team feels so much bigger than the sum of its parts — they’re individually brilliant but highly collaborative. We genuinely care to solve the problems we’re tackling and have a track record of getting stuff done. I hate BS and so we avoid it. And we’re working towards a mission that’s so rewarding, work feels like play!
Come join us if you’re looking for opportunities. We are especially looking for super ML & data engineers and data products folks who want to use their superpowers to improve the way health care is delivered. Email careers AT @bayesianhealth.com
Suchi Saria
@suchisaria
Feel free to RT👆🏽 if you feel you have a following of amazing AI/ML folks who'd love to work in a mission oriented kicka** company. Thanks!
-
Suchi Saria
@suchisaria
Served as a presenter/discussant at a brilliant White House discussion earlier today with heads from the FDA, ONC, HHS, VA and 30-40 representatives from different agencies. Q&A was focused on (a) understanding current state (b) will it raise cost (c) consent (d) regulation and infrastructure for reg (e) patient empowerment. Very level headed discussion.
Interest in AI safety and reg is at an all time high. However, we need more individuals with deep knowledge of AI, experience translating solutions to the real world and regulation. -
Suchi Saria
@suchisaria
My feed is all #ai AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI …
-
Suchi Saria
@suchisaria
If you’re looking for LLM project ideas,👇 is a superb one — doesn’t work yet but likely doable based on performance for code completion and would be a very useful app if it worked.
AK
@_akhaliq
Can LLM Already Serve as A Database Interface? A BIg Bench for Large-Scale Database Grounded Text-to-SQLs
abs: arxiv.org/abs/2305.03111 -
Suchi Saria reposted
chilconference
@CHILconference
We can't wait to hear the exciting talks on the state of ML for health from @suchisaria, @kdpsinghlab, and @drnigam on Saturday of the conference (June 24)! Afterwards, @byron_c_wallace will be moderating a panel on LLMs like GPT-4! Register now at www.chilconference.org/register.html
-
Suchi Saria reposted
Brittany Trang
@brittanytrang
My recap of @caseymross' conversation with @drnigam, @suchisaria and @connorlandgraf about #ai at the @statnews Breakthrough Summit:
We're thinking about AI all wrong, especially when we reduce it to just #chatgpt:
www.statnews.com/2023/05/05/chatgpt-health-care-ai-hospit… -
Suchi Saria
@suchisaria
Google "We Have No Moat, And Neither Does OpenAI". Fascinating article on AI strategy 👇
It very clearly responds to many commonly held misunderstandings about LLM and GPT tech development progress.
www.semianalysis.com/p/google-we-have-no-moat-and-neither
Suchi Saria
@suchisaria
1/ Development is going to be controlled by the big 3-4 tech cos. ❌
2/ It’s prohibitively expensive to train good models ❌
3/ “The value of owning the ecosystem cannot be overstated.”
Suchi Saria
@suchisaria
4/ Slightly more subtle but an OpenAI style “LLM for Domain X” company is a bad plan. On the flip side, companies with deep tech expertise that can unbundle and rebundle this tech in a purposeful way for building great user experiences and solving core user problems will benefit from this rapid development! (Self-serving: delighted to see the arguments so clearly laid out; predicted that this is the way this tech would go and @bayesianhealth’s last 12 month’s trajectory/strategy is built around this.)
Great read via @nathanbenaich
Thoughts? -
Suchi Saria reposted
STAT
@statnews
AI in Health Care: What are the Risks?
The ability of AI models to answer medical questions & quickly summarize information has awakened the world to its potential. But they also commit grievous errors & make things up.
Ft. @drnigam @suchisaria @connorlandgraf
#statbreakthrough -
Suchi Saria
@suchisaria
Breaking: Biden-Harris Administration Announces New Actions to Promote Responsible AI Innovation that Protects Americans’ Rights and Safety 🧵
www.whitehouse.gov/briefing-room/statements-releases/2023…
Suchi Saria
@suchisaria
New funding for fundamental research — the @NSF (National Science Foundation) is announcing $140 million in funding to launch seven new National AI Research Institutes. This investment will bring the total number of Institutes to 25 across the country,
It’s a fraction of what industry is spending but good to see nonetheless.
Suchi Saria
@suchisaria
Focus on evaluations — the Administration is announcing an independent commitment from leading AI developers, including Anthropic, Google, Hugging Face, Microsoft, NVIDIA, OpenAI, and Stability AI, to participate in a public evaluation of AI systems, consistent with responsible disclosure principles—on an evaluation platform developed by Scale AI—at the AI Village at DEFCON 31.
Suchi Saria
@suchisaria
Clear guidelines and policies drafted by the OMB for empowering responsible use of AI by the government.
Great to see! -
Suchi Saria reposted
Andrej Karpathy
@karpathy
Excellent TED talk from Sal Khan:
- many inspiring examples of GPTs finetuned into socratic tutors, assisting without giving away answers.
- none of it "out of the box", requires prompt engineering, finetuning, data collection, iteration.
- sense of barely scratching the surface.
Salman Khan
@salkhanacademy
What do you think? www.ted.com/talks/sal_khan_the_amazing_ai_super_tutor_for…
-
Suchi Saria
@suchisaria
This is an extremely cool and phenomenally impactful dataset. The goal here is to make sense of the babies’ crying sounds. By doing so, we both understand the “language of babies” and how they communicate but also identify cries that maybe indicative of poor health outcomes.
Please RT if a lot of people doing ML follow you.
Charles C Onu
@onucharlesc
The competition is now live! We have released a first-of-its-kind dataset for this:
👶 26,093 cry sounds
⏺ 1,300 recordings
👶 786 infants
📶 6.5 hours of cry signals
Participate : docs.google.com/forms/d/e/1FAIpQLSdo8Gp352VsNgBYN1kd7R_KI…
Paper: arxiv.org/abs/2305.00969 -
Suchi Saria
@suchisaria
Excellent news! 👇 is happening.
Working through the plans. More to come.
If you have ideas on who do you find to be the most compelling on either side, would love ideas.
We will also be collecting a list of key ideas/quotes/POV from the “extreme left” to the “extreme right” on the topic and flesh these ideas out live. The goal of our discussion will be to get to an understanding of the points of agreement and disagreement, and a way forward!
Suchi Saria
@suchisaria
. @ylecun and @geoffreyhinton … it would be very useful to find time to discuss/debate AI safety. I was very surprised to see Geoff’s remarks in NYT today and it would be a really productive discussion amidst the current discourse/chaos.
If relevant, I can bring the FDA/regulatory point of view based on my work there on ensuring AI safety so we can draw takeaways other sectors.
@kchonyc @BeEngelhardt @EmmaBrunskill should we do something like this at ICML? Can do part virtual if Geoff isn’t planning to join in person.
All, RT or add ideas if you’d like to see this happen.
Suchi Saria
@suchisaria
@shakir_za reminds me of the ICML debate several years ago we had on should AI be regulated!
-
Suchi Saria
@suchisaria
. @ylecun and @geoffreyhinton … it would be very useful to find time to discuss/debate AI safety. I was very surprised to see Geoff’s remarks in NYT today and it would be a really productive discussion amidst the current discourse/chaos.
If relevant, I can bring the FDA/regulatory point of view based on my work there on ensuring AI safety so we can draw takeaways other sectors.
@kchonyc @BeEngelhardt @EmmaBrunskill should we do something like this at ICML? Can do part virtual if Geoff isn’t planning to join in person.
All, RT or add ideas if you’d like to see this happen.
Suchi Saria
@suchisaria
Suchi Saria
@suchisaria
Excellent news! 👇 is happening.
Working through the plans. More to come.
If you have ideas on who do you find to be the most compelling on either side, would love ideas.
We will also be collecting a list of key ideas/quotes/POV from the “extreme left” to the “extreme right” on the topic and flesh these ideas out live. The goal of our discussion will be to get to an understanding of the points of agreement and disagreement, and a way forward!
Suchi Saria
@suchisaria
. @ylecun and @geoffreyhinton … it would be very useful to find time to discuss/debate AI safety. I was very surprised to see Geoff’s remarks in NYT today and it would be a really productive discussion amidst the current discourse/chaos.
If relevant, I can bring the FDA/regulatory point of view based on my work there on ensuring AI safety so we can draw takeaways other sectors.
@kchonyc @BeEngelhardt @EmmaBrunskill should we do something like this at ICML? Can do part virtual if Geoff isn’t planning to join in person.
All, RT or add ideas if you’d like to see this happen. -
Suchi Saria reposted
Brittany Trang
@brittanytrang
You do not want to miss out on Casey asking questions!! Always a good time (unless you're the one answering the questions...)
The Qs will be especially good for his "risks of AI in healthcare" session on Thursday, ft. @suchisaria, @drnigam and @connorlandgraf -
Suchi Saria
@suchisaria
Wow, what’s not to love — the GREAT CATSPY! 🐈⬛ 🍾🪩
Made entirely by #ai text-to-video!
Nathaniel Whittemore
@nlw
This entire video was made by a guy typing some words into a box (a bunch of times). The degree to which that this will unleash human creativity is stunning.
This media couldn't be displayed. See it on the original post. -
Suchi Saria
@suchisaria
The message in @YejinChoinka’s talk is right on.
Simplifying complex ideas is often hard and she does a super job at that.
For those who are newer to these ideas, did it make sense? Any questions?
Gary Marcus
@GaryMarcus
Don’t believe the hype. Instead watch this marvelous TED talk by @YejinChoinka. #ai #ml
www.youtube.com/watch?v=SvBR0OGT5VI -
Suchi Saria reposted
Sharon Zhou
@realSharonZhou
I’m super excited to announce @LaminiAI, the LLM engine that gives every developer the superpowers that took the world from GPT-3 to ChatGPT! We make it easy to rapidly train custom LLMs from @OpenAI @EleutherAI @Cerebras @Databricks @HuggingFace @Meta lamini.ai/blog/introducing-lamini 🧵
-
Suchi Saria
@suchisaria
Everybody wants a piece of the action. Excited for the acceleration.
** Deep understanding of problem + deep tech + focus + great partnership culture ** is winning combo here.
Recurring Q I heard on the road last 5 weeks — we’re on the outside & what’s the best place to start. Find good partners. Calling Microsoft will feel safe but find focused partners.
Aakash Gupta
@aakashgupta
The week is just over halfway through, and we've seen 8 huge announcements in AI:
Apple Enters AI
Replit a Unicorn
Mobile AutoGPT
Yelp AI in Search
TikTok AI Avatars
Secrets in ChatGPT
Boston Dynamics GPT Robot
Hugging Face ChatGPT Alternative
Here's what you need to know 👇 -
Suchi Saria
@suchisaria
Impressive that this could be built so quickly.
My guess is — a much less uncanny, more natural version will exist before the end of the year!
This could be such the perfect companion for venting when you need to let off some steam :-)
Chris Frantz
@frantzfries
Not to be too much of an AI Doomer but I just had a FaceTime call with a near realtime ChatGPT-powered AI avatar and it was…surprisingly human.
Anyway, I recorded it.
This is the actual response speed, it’s very fast.
(You can also hear me kind of lose it towards the end) -
Suchi Saria
@suchisaria
Listening to the superbly accomplished Mary J. Blige on the Colbert show. She speaks about how she wrote this book “Mary Can!” bcoz she was told “she can’t”!
I’ve heard so many accomplished women start from being told “you can’t”.
I wonder if this is common for men too? -
Suchi Saria
@suchisaria
Just came across @TIME 100.
Delighted to see @mmitchell_ai alongside @sama and @iamsrk on it! Congrats! Both AI and India are represented well!
Also, was surprised to see Elon getting trolled — “What’s the opposite of progress? Elon Musk 2023”
time.com/collection/100-most-influential-people-2023/ -
Suchi Saria
@suchisaria
Proclamations are easy,
results are HARD.
Speculating is easy,
building is HARD.
Critiquing is easy,
problem solving is HARD.
Barely-baked prototypes are easy, shipping validated solutions’ V HARD.
Hype cycles occur when 99.9999% of the activity is focused on the former, by folks who aren’t learning by doing.
Let’s solve, build, test, learn.
Suchi Saria
@suchisaria
Ans if you’re a data engineer or ML-focused problem solver seeking good company, join me here at @bayesianhealth
We are backed by reality and results :-) -
Suchi Saria
@suchisaria
Excited to see @bayesianhealth on @Forbes AI 50!
It’s been quite the year in AI with 100s/1000s of new ventures focusing in the space. Great to see other amazing peers and friends also recognized!
Suchi Saria
@suchisaria
We are hiring in data engineering and ML. It’s a solid team, great culture, fun/breakthrough tech problems, and a mission that makes work not feel like work.
If interested, email careers AT bayesianhealth DOT com or apply on website. -
Suchi Saria reposted
Jim Fan
@DrJimFan
Below is Steve Jobs announcing a new product, “WiFi-connected socks” in his eloquent voice. I’m sold.
The Black Mirror episode “Be Right Back” describes a technology that revives people from their digital footprint. Where we are today:
- LLMs prompted by past chat logs can do a great job. For example, @character_ai can simulate various persona for you to chat with.
- Highly realistic, personalized speech synthesis has made big progress, and will reach perfection in 1-2 years at most. @elevenlabs
- Face simulation with natural emotions is much harder. @UnrealEngine introduces a beta service called “MetaHuman” that may fill the gap.
- Physical robot of your loved ones is completely out of reach for today’s tech. Even if it’s doable, there would be huge legal and moral red flags.
Demo made using @forever_voices -
Suchi Saria
@suchisaria
This👇 is something we should be quite worried about (a great deal more than AGI becoming sentient and overpowering humans).
Are there groups designing guardrails? Perhaps we should be discussing exploratory regulation? Since this is such a fast evolving field, should a centrally coordinated group be crowdsourcing ideas and prioritizing mitigation strategies?
It’s unclear if we are doing enough.
Jeremy Kahn
@jeremyakahn
The 2024 US presidential election could be even crazier than any we’ve seen before due to generative A.I. Thanks for sharing your thoughts @chrismeserole @GaryMarcus @SandraWachter5 fortune.com/2023/04/08/ai-chatgpt-dalle-voice-cloning-202…
-
Suchi Saria
@suchisaria
Taking a traditional healthtech co & getting it to build new advanced tech like AI-led products is surprisingly hard.
Building a culture that deeply understands both domain + tech & really sees the picture end-to-end is non trivial. You can’t just hire experts in each and put them in a room. You’ll often end up with the wrong hires and much is lost in translation.
It’s interesting to see when incumbent advantage wins…
Joshua Liu
@joshuapliu
Yup
Also - incumbents with Brand, Trust & Distribution that had compounded over years will win in the Healthcare Provider space - whether it’s AI or some other new Tech
Case in point: Nuance (Microsoft) and now adding OpenAI capabilities. Nuance has years of Provider relationship and huge distribution edge now -
Suchi Saria
@suchisaria
Historically, creating video games was very expensive. This market is being disrupted significantly.
- AI tech 4 creating 3D worlds is advancing super fast
- You can plug🔌 into game market places 4 distribution
I can envision 10yr olds designing popular video games that make💰
Blockade Labs
@BlockadeLabs
REEEMIXXXXX! 📢📢📢
Infinite 360° worlds just got easier to fine tune. Transform anything - including visual style, time of day, & weather while keeping the structure intact!
LIVE NOW: skybox.blockadelabs.com
#skyboxlab #texttoskybox #aiart #gamedev #aitools -
Suchi Saria
@suchisaria
I’ve worked in AI for 20+ yrs & healthcare for 12. HC is super hard. I’d bet a lot more on a company that understands AI & HC as opposed to a team w/ deep AI but no HC background. The latter is a much harder field to crack into.
For teams w/ tech only experience that do succeed, it’s often when they’re building a better mousetrap in a mature market.
Curious to hear your experience.
Jack Needham
@JNeedem
there a 1000 AI companies starting to tackle administrative healthcare processes
from clinical notes to patient history to prior-auth
but it feels like the LLM is the commodity, despite its complexity
so what are the long-term moats?
it seems like this could be a race to the bottom for the startups and the incumbents entrench themselves more via distribution
creds @healthapiguy bringing this front of mind today -
Suchi Saria
@suchisaria
I’ve been using the @perplexity_ai app on my phone for search and it’s been a great experience. Rarely go to chrome now.
-
Suchi Saria
@suchisaria
Great read by @rachelmetz
Anthropomorphizing AI is a great way to capture people’s imagination around AI and make a pile of math exciting and accessible!
But it’s also likely the root cause of disagreements around the right way to approach AI safety.
www.bloomberg.com/news/newsletters/2023-04-03/chatgpt-bin… -
Suchi Saria
@suchisaria
I love articles that give a sneak peek into how people operate. This was a good one by @CadeMetz on @sama
Was surprised to learn that Sam has minimal to no equity in OpenAI. Was also surprised to read Paul Graham’s speculation on why - potentially power.
Sharon Goldman
@sharongoldman
But on the other hand, now @CadeMetz dives straight into the mix in his profile of @sama yesterday. www.nytimes.com/2023/03/31/technology/sam-altman-open-ai-…
Suchi Saria
@suchisaria
Here is a poll based on this article. Finding it fascinating to see responses. Please RT. Will publish results once poll ends.
Suchi Saria
@suchisaria
@CadeMetz in his NYT piece writes on the future of AI: “Some believe it [AI] will deliver a utopia where everyone has all the time and money ever needed. Others believe it could destroy humanity. Still others spend much of their time arguing that the technology is never as powerful as everyone says it is, insisting that neither nirvana nor doomsday is as close as it might seem.”
Which camp are you in? Pick the poll that best describes you. Pls RT. -
Suchi Saria
@suchisaria
@CadeMetz in his NYT piece writes on the future of AI: “Some believe it [AI] will deliver a utopia where everyone has all the time and money ever needed. Others believe it could destroy humanity. Still others spend much of their time arguing that the technology is never as powerful as everyone says it is, insisting that neither nirvana nor doomsday is as close as it might seem.”
Which camp are you in? Pick the poll that best describes you. Pls RT.Post not available here, view on X
Suchi Saria
@suchisaria
If you haven’t published but are broadly interested in AI, which camp do you fall in?
Suchi Saria
@suchisaria
Will publish data for both polls when the poll ends. Please don’t vote in the wrong poll simply to see results. Thank you.
-
Suchi Saria
@suchisaria
Nice work @mdredze @gideonmann and the Hopkins and Bloomberg team. Excited to see more domain specific models and uncover where gaps lie in real-world challenging tasks.
We’ll learn a lot from domain experts pushing this tech (compared w/ the wild speculation on Twitter).
elvis
@omarsar0
BloombergGPT is a new LLM for finance.
It's a 50 billion parameter language model trained on financial data.
Claims the largest domain-specific dataset yet with 363 billion tokens... further augmented with 345 billion tokens from general purpose datasets.
Outperforms existing models on financial tasks while not sacrificing performance on general LLM benchmarks.
arxiv.org/abs/2303.17564v1 -
Suchi Saria reposted
William Green
@WilliamGreen72
Jeff Bezos on the power of a long time horizon: "If everything you do needs to work on a 3-year time horizon, then you're competing against a lot of people. But if you're willing to invest on a 7-year time horizon, you're now competing against a fraction of those people..."
-
Suchi Saria
@suchisaria
This photo feels like a repeat of 2015!
In 2015, Deepmind was showing exciting progress with DeepRL. Elon was worried about AI progress being concentrated at Deepmind leaving him w/o any access to it. What followed was many news headlines from Elon calling out AI risk. And, he ended up getting together with Sam A, Greg, Ilya and others to form OpenAI.
Fast forward to 2018, AI progress seemingly “slowed down” (read: no good demos), and he moved focus to AI @ Tesla and left OpenAI.
It’s 2023, progress is back up again and OpenAI is closed, and feels like history is repeating itself. Elon is back discussing AI risks. Will there be funding for a new OpenAI competitor? Will we actually do something more tangible (besides starting another AI company) around responsible AI deployment and adoption? -
Suchi Saria reposted
Santiago
@svpino
You can now clone ChatGPT!
OpenAI didn't open-source its models, so we don't know much behind the scenes.
But the first complete end-to-end model pipeline was just released, and it's the most practical open-source project resembling ChatGPT.
Here are the details: -
Suchi Saria
@suchisaria
Thinking of having this as a means of transport in New York City. #mars2023
-
Suchi Saria
@suchisaria
Felt so honored and lucky to be funded by Gordon Moore. He is a legend in our field and he funded some of the most innovative work done in our lab.
- Early funding from him went towards the first ever demonstrations of automated measurements to improve hospital safety: pubmed.ncbi.nlm.nih.gov/24303253/
- Next we showed the first feasibility papers for early detection of patient harms like sepsis from AI & multimodal data: www.science.org/doi/10.1126/scitranslmed.aab3719
- Next, we began working on moving AI from the lab into the real world, leading to these first ever large prospective studies of multimodal AI improving patient outcomes: www.nature.com/articles/s41591-022-01894-0
While he is most known for his work in core computing, through research funding, he enabled transformative advances in many fields. Many government agencies thought of our work in health AI in 2013 as too new and risky but GBMF championed it, and it’s now given rise to whole new lines of work!
🙏 RIP
Laura Landro
@LauraALandro
RIP Gordon Moore..that fortune he built and the vision of his wife Betty Gordon Moore funded so many interesting and worthy projects to improve patient experience and outcomes in health care via the @MooreFound www.nytimes.com/2023/03/24/obituaries/gordon-moore-dead.html
Suchi Saria
@suchisaria
Gordon Moore was the Moore of Moore’s law & the co-founder of Intel.
-
Suchi Saria
@suchisaria
Anyone interested in faculty positions in machine learning at the associate professor level at a top university in Australia?
If so, respond here with your CV and happy to make the connection. RT if you have a big research/faculty following. Tx! -
Suchi Saria reposted
Denis Wirtz
@deniswirtz
Hopkins is recruiting six Bloomberg Distinguished Professors and six TT Assistant Professors in neuroX.
Interested? Contact Prof. @triciajanak and Prof. @rhuganir. -
Suchi Saria
@suchisaria
Riveting - engineering biology to produce a baby from two males!
While we are all getting excited about developments in AI, our scientific advances in genetics/genomics are advancing at a pretty cool pace too.
Great time to be a scientist/builder.
Steven Salzberg 💙💛
@StevenSalzberg1
Two males can now have a baby - if they're mice. It won't be long before two human males can do this. I think it's pretty cool, though I can imagine that it will drive some folks crazy 😅 www.forbes.com/sites/stevensalzberg/2023/03/20/these-mice…
-
Suchi Saria
@suchisaria
Would love more nominations. DMs will do too!
cc @IAmSamFin @atulbutte @Rbaltman @EricTopol @_asubbaswamy @royjamesadams @kdpsinghlab @zakkohane @oziadias @DrJimFan @DrewPrinster @coverbytaylor @marksendak @GreeneScientist @ThomasFuchsAI @kyeshwant
Suchi Saria
@suchisaria
Friends, I’m putting together an invited year in review for conference on health, inference & learning (CHIL) www.chilconference.org
Was asked to highlight best of 2020-2022 related to Health AI, informatics, inference & learning.
Would LOVE pointers to research you thought helped *significantly advance the field*. For each ref, please add what is the key contribution & why it’s significant 🙏 -
Suchi Saria
@suchisaria
Will SMB behavior around banking change post SVB?
Going forward, will we have two classes of banks: (a) too big to fail (eg BOA, JPM) and (b) service focused banks (eg FRB)
Individuals leave their large pools of capital in one of the big banks but keep an account with limited capital in another smaller bank like FRB that knows how to delight customers? Does this give them enough capital to excel at operating?
I hated banking with @BankofAmerica and am regretting needing to go back.
Anshu Sharma 🌶
@anshublog
I bank with @firstrepublic ‘cos when I was without a full-time job in ‘13, they were the only one who looked at me as a person - analyzed my assets and income - and gave me a mortgage.
I have started 3 companies, all bank with them.
They do the work of a bank.
The work of a bank is to analyze the risk, make loans, keep the customer, and provide great service. The software is dated but the people are awesome.
I want my bank to pick up my phone. For 9 years, they have always picked up the phone - including this weekend. -
Suchi Saria
@suchisaria
Friends, I’m putting together an invited year in review for conference on health, inference & learning (CHIL) www.chilconference.org
Was asked to highlight best of 2020-2022 related to Health AI, informatics, inference & learning.
Would LOVE pointers to research you thought helped *significantly advance the field*. For each ref, please add what is the key contribution & why it’s significant 🙏Post not available here, view on X
Suchi Saria
@suchisaria
Self-nominations are absolutely welcome. You’ve worked hard so be proud!
If you’re in industry and you’ve done stellar work around studying real-world deployments, even if it’s only in abstract form, would love to hear.
This is the AGE of health + medicine being transformed by smart software!
Suchi Saria
@suchisaria
Please RT if you work in this area so we can be as inclusive as possible. Thank you!
-
Suchi Saria
@suchisaria
Things I have experienced while growing my startup @bayesianhealth
- a pandemic 🦠
- a war 🇺🇦
- a recession 📉
And, today
- a bank failure! 🏦
What doesn’t kill you makes you 💪🤷♂️ What a ride!
Sending 🤗 to all who were impacted by SVB over the last 36 hours. -
Suchi Saria
@suchisaria
Giving an invited talk at the @AHRQNews meeting shortly on Racial Bias and Algorithms. 🧵 on some interesting takeaways! 1/
Suchi Saria
@suchisaria
Examples of published papers with mitigation strategies used to date for reducing algorithmic bias …
Suchi Saria
@suchisaria
Presented an end-to-end checklist for bias detection. The intended users are individuals looking to deploy/use (eg health system operator) algorithms; this helps the decision maker assess whether there is bias and what kind that materially impacts outcomes in their environment.
-
Suchi Saria
@suchisaria
I realize we are all very obsessively discussing AGI but it’s exciting to see applications of AI serving as a co-pilot to doctors and nurses and saving lives!
More about it in @WSJ today covering our work at @bayesianhealth www.wsj.com/articles/how-doctors-use-ai-to-help-diagnose-…
We’re growing. Join us! -
Suchi Saria
@suchisaria
Having spent the last 3 years leading research with the FDA on tools for AI safety, AI as a field can learn a lot about ways to ensure safety from healthcare/medical devices — lessons on what works well as well as what not to do.
Self regulation is likely not enough.
www.wsj.com/articles/timnit-gebru-is-calling-attention-to…
Suchi Saria
@suchisaria
We debated this on stage at ICML several years ago w/ @shakir_za @isbellHFh @mkearnsupenn Repeat?
Techies are typically opposed to regulation b/c it slows down innovation.
In healthcare, FDA has expanded oversight & my bet is that it accelerates AI adoption
icml.cc/virtual/2021/social/12458 -
Suchi Saria
@suchisaria
While Elon's brilliant, this feels like the showman side of him www.theinformation.com/articles/fighting-woke-ai-musk-rec…
During the last "AI is hot" wave, he wanted to be in on the action, funded OpenAI ...but didn't have the patience to stick it out & cut his ties too early.
Seems like he's doing it again.
Suchi Saria
@suchisaria
@sharongoldman perhaps Elon might share his motivations with you in more detail if you end up pursuing this story
-
Suchi Saria reposted
Adam D'Angelo
@adamdangelo
Possibly the worst takeaway you could have from web3 being overhyped and then failing to deliver is that the optimism around AI means it will fail to deliver. If there was ever a case for first principles thinking or digging into details beating pattern matching this is it.
-
Suchi Saria
@suchisaria
I'll be on NPR tomorrow @ 10am ET! If there is anything you want to hear about AI or Health AI, comment below!
The live interview is on NPR's 1A show (www.npr.org/podcasts/510316/1a). I'm told guests will have the opportunity to ask questions so bring your toughest/most fun ones! -
Suchi Saria
@suchisaria
Very nice paper by @HsseinMzannar @erichorvitz—measures how coding workflows change w/ a copilot
Last yr, we published on what’s needed to build successful copilots 4 docs & nurses: www.nature.com/articles/s41746-022-00597-7
Exciting to see increased focus on understanding AI-driven augmentation!
Hussein Mozannar
@HsseinMzannar
As Copilot becomes more popular, we need to understand how programmers interact with it. We built a model of interaction between Copilot and Programmers named 'CUPS' and predict programmer behavior in our latest paper arxiv.org/abs/2210.14306
-
Suchi Saria
@suchisaria
Need your help. Responses to any are welcome!🤗
1) How would you interview for whether someone is going to be fast at learning new things?
2) Independently, how do you evaluate for whether they’re a critical thinker?
3) How do you evaluate 4 whether they’d be solid @ product?
Suchi Saria
@suchisaria
I’ve easily interviewed 1000+ people —feels like a really, really tricky skill to be able to spot great and productive people.
Still refining my instinct and looking to learn if you have suggestions … -
Suchi Saria
@suchisaria
I was interviewing with a clever reporter about AI & ChatGPT which got me wondering:
- we were good at building symbolic reasoning systems (SRS) for logical entailment and inference
- to avoid hallucination in summarization tasks, can we combine a generative AI approach with a SRS? Have you seen practical demonstrations of this?
cc @tdietterich @ylecun @pmddomingos -
Suchi Saria
@suchisaria
🤯😆😂
Except 👇 is also what #ai research feels like!!
Suchi Saria
@suchisaria
Every new wave comes with new talent that rediscovers several foundational ideas under a new name and framework! :-)
-
Suchi Saria
@suchisaria
This is a 🧵 on MLOps / AI safety.
I’d love pointers for closing gaps in my understanding.
Starting w/ background first 👇Post not available here, view on X
Suchi Saria
@suchisaria
2/ I see tools/platforms for visualizing model performance (eg Grafana). Any solutions that help you discover shifts, identify cause, and/or know when to retrain and where to focus retraining?
Many open research problems here but what if any industry tooling are available here?
Suchi Saria
@suchisaria
Please RT if you have peers/follows who work in MLOps or are in academia and research ML/AI infra for improving real-world development and deployments. Huge thanks!
-
Suchi Saria
@suchisaria
Watching Pathaan in a theatre! 😎
Haven’t watched a Hindi movie in a theatre in the US. Feels like I’m back in India. Loving it! -
Suchi Saria
@suchisaria
AI excitement is back and at an all time high!
Having seen a few cycles of ups & downs with AI over the last 15 years, a few takeaways for investors and those newer to the field. 🧵-
Suchi Saria
@suchisaria
1/ Many ask if all this recent progress is new and surprising.
Exciting - yes!
Overnight success - no!
AI’s been making very consistent progress over the last decade, perhaps two.Post not available here, view on X
Suchi Saria
@suchisaria
6/ This time around Twitter & LI makes way for even more pundits and talking heads.
Would love to avoid an AI winter because of hyped investments that don’t pan out.
Let’s innovate responsibly! We don’t want AI equivalents of Holmes & SBF!
Other ideas? Pitfalls to avoid?
Suchi Saria
@suchisaria
If you’re in AI and looking to work with a 🚀 team with breadth and depth in AI and healthcare, check out @bayesianhealth
If you’re a team looking for funding for your own company, check out @aixventureshq
Or DM if you have questions. -
Suchi Saria
@suchisaria
If you’re in clinical informatics & want to design software that saves lives, we’re 📈. Come join us!
We’ve got a super🚀team building AI for augmenting docs+nurses. It’s already shown game changing results: bayesianhealth.com
RT if you’ve peers/students looking for opp
Suchi Saria
@suchisaria
Reqs:
💥Curious, self-starter, & learns new ideas quickly
🛠 Has experience building a clinical informatics solution
💻Knows SQL; can code, test, map clinical terms to code for data extraction
🍻Collaborates & writes well
☺️ Is kind
Clinical training is a plus but not required
Suchi Saria
@suchisaria
Experience collaborating w/ clinicians is necessary.
Role is fully remote. Ideal is part time but FT will be considered. Must have authorization to work in the US.
Send resume including description of one or two past CI projects to careers__at__bayesianhealth.com -
Suchi Saria
@suchisaria
If you’re in AI and want to make a massive impact & save lives, we’re 📈. Come join us!
We’ve got a super team building AI for augmenting docs+nurses. It’s already shown game changing results: www.bayesianhealth.com
We work hard & play hard! RT if you’ve friends looking 4 opp.
Suchi Saria
@suchisaria
If you have questions, DM me or ask here! Can’t wait to see your CVs and possibly work with you!
Suchi Saria
@suchisaria
If your background is in clinical informatics or product, we are looking for hires there too!
-
Suchi Saria
@suchisaria
Awesome that docs *want* to use AI & ChatGPT 4 medical use cases. Bodes well for improving practice & outcomes!
But 😢 terrible that they’re extrapolating what AI can/can’t do in medicine from it 👇
Curt Langlotz
@curtlanglotz
"Then I asked #chatgpt the reference for why costochondritis is more likely when taking oral contraceptive pills. It came up with a study in the European Journal of Internal Medicine. I look that up, and it's made up. That's not a real paper." www.medpagetoday.com/opinion/faustfiles/102723
Suchi Saria
@suchisaria
ChatGPT was designed to predict words & mimic human dialogue. It wasn’t designed to be factually correct. And it’s not indicated for medical use.
We are, however, seeing absolutely amazing progress in HealthAI outside of ChatGPT! As an example, check out @bayesianhealth -
Suchi Saria
@suchisaria
Who are your favorite people working in Health AI? I mean those who understand ML and, more importantly, how to bring these approaches to the REAL WORLD at scale?
Please RT if you’re connected with real-world AI and deployment focused groups.
Suchi Saria
@suchisaria
Self-promotion is welcome! Please describe your experience.
Suchi Saria
@suchisaria
If you’re naming an individual or group, please describe work from them that’s inspiring it. Thanks!
-
Suchi Saria
@suchisaria
Many AI predication threads are hyped & misinformed but really enjoyed👇
Highlights:
• CoPilot/Augmented Intelligence for X: these startups are a GREAT oppty to build breakthrough products, in important markets, & with a moat
🧵 1/
sridhar
@RamaswmySridhar
2/ Let's first dive into what the industry looks like today.
We see 5 main categories:
⭐ Foundation Model Players
⭐ AI Frontend Startups
⭐ Copilot/LLM for X Companies
⭐ Tooling Companies
⭐ Big Compute CloudsPost not available here, view on X
Suchi Saria
@suchisaria
• @aixventureshq has some of the most well known #ai experts w/ deep experience building research-based category-leading AI products:
🦾 AI + manufacturing @pabbeel @CovariantAI
🔍 AI + search @RichardSocher @YouSearchEngine
⛑ AI + health moi @bayesianhealth
3/
Suchi Saria
@suchisaria
+ nlp guru @chrmanning and kaggle founder @antgoldbloom
Together, the AIX team has advised many of the best AI startups across many domains.
If you’re looking for funding or looking to work with us or one of our portco startups, DM me or @aixventureshq -
Suchi Saria reposted
François Chollet
@fchollet
"Narratives based on zero data are accepted as self-evident"
Jimmy Baker
@JimmyBa62254692
ChatGPT already outperforms most doctors in the USMLE. It will only be a short matter of time when we replace them with AI to save lives and improved outcomes. This is a trillion dollar industry.
-
Suchi Saria
@suchisaria
8 years ago when I first started learning causal inference, I used 👇 book draft heavily and read it alongside several others including Pearl’s book. Great to see it published and echoing my endorsement again!
Rob Calver
@RobCalver5
"Anyone interested in data science and machine learning should read this book. To form valid models of the world from data, understanding causal inference is critical. They’ve provided a comprehensive and accessible introduction to the topic."
- Suchi Saria @suchisaria -
Suchi Saria reposted
Eric Zhu
@ericzhu
People are calling 2023 the year of AI which is really exciting!
AI is calling 2023 “The last year of people” which is a little disconcerting. -
Suchi Saria reposted
Marc Andreessen 🇺🇸
@pmarca
We are just entering an AI-powered golden age of writing, art, music, software, and science. It's going to be glorious. World historical.
-
Suchi Saria
@suchisaria
I’ve been part of many hiring committees (and tried to remove biases) & unfortunately a lot of this resonates. For those on the academic job market, this thread might be helpful.
Oversupply of highly qualified candidates and scarcity of roles leads to all this!
Paul Schofield
@pschofie79
Academic job seekers often think that because they didn't get a job one year, or didn't get a particular job, there's something they should change in their application. That's often a mistake. Here are some reasons you maybe didn't get the job that are beyond your control(thread)
-
Suchi Saria
@suchisaria
Sunrise this morning in the Atacama desert 🌵🦙 with hot geysers in the foreground.
Has been a very experience-filled year & preparing for goodbyes in a fitting fashion! -
Suchi Saria
@suchisaria
Love that our work in using #ai to improving sepsis outcomes (led by @bayesianhealth ) is called out! So cool! And several other incredible inventions you’ll recognize including the covid dashboard!
Johns Hopkins University
@JohnsHopkins
2022 was a big year for Johns Hopkins University! See how students, faculty and staff made an impact.
This media couldn't be displayed. See it on the original post. -
Suchi Saria
@suchisaria
There is no shortage of podcasts but this one should be solid — it will further help bridge the gap b/w AI and it’s use in health care!!!
Looking forward to hearing @arjunmanrai and @AndrewLBeam as hosts lead the way.
NEJM AI
@NEJM_AI
NEJM AI Grand Rounds, a new podcast from NEJM Group exploring #artificialintelligence and its application to medicine, is coming later this month. Listen to a preview and subscribe: ai-podcast.nejm.org #aiinmedicine
-
Suchi Saria
@suchisaria
One of my favorite ChatGPT threads.
Arvind Narayanan
@random_walker
ChatGPT can only multiply small numbers correctly. If you ask it to explain step by step, it generates nonsense. But you can just describe the multiplication algorithm in words and it will execute it flawlessly — except call it smth like "snorp" so it doesn't use its wrong algo!
-
Suchi Saria
@suchisaria
Slides from my invited talk at #neurips2022 describing the latest and greatest on approaches for ensuring safety and reliability as we move tools from the lab to the real-world: www.dropbox.com/s/28wn2pv81j5tafp/NeurIPS-robustness-work…
With @_asubbaswamy @DrewPrinster @anqi_liu33
Enjoy! -
Suchi Saria
@suchisaria
ChatGPT is super cool 😎 — all the demos I’m seeing are of storytelling or fact extraction around a concept and told/organized in a particular style. Can get away w/o language understanding.
This work below sounds even harder as it seems like it would require reasoning?!
Google DeepMind
@GoogleDeepMind
Could AI help people with a wide range of views find agreement?
Our team fine-tuned a 70 billion parameter language model to generate statements that could bring consensus among groups with diverse opinions: dpmd.ai/3XR3Bm3 #neurips2022 -
Suchi Saria
@suchisaria
Machine learning predictive uncertainty estimates are often unreliable—data shift makes things worse!
How can you audit the uncertainty of an ML prediction, even with biased data?
A 🧵 w/ @DrewPrinster on the JAWS approach in #neurips2022 paper w/ fab @anqi_liu33 @DrewPrinster
Drew Prinster
@DrewPrinster
If you’re at #neurips2022 come see our poster Hall J #941 TODAY (11/29) 4-6pm!
With @anqi_liu33 @suchisaria, we present “JAWS: Auditing Predictive Uncertainty Under Covariate Shift”
Virtual talk: neurips.cc/virtual/2022/poster/52881
Paper: arxiv.org/abs/2207.10716v2Post not available here, view on X
Suchi Saria
@suchisaria
(Contribution 3) Error assessment: General approach to repurposing predictive interval generating methods to error assessment, based on user’s error criterion of interest (for instance, the user wants the chance of false negative or missed diagnosis to be < 5%)
Suchi Saria
@suchisaria
(Contribution 4) Experiments: Demonstrate superior empirical performance of JAWS over conformal and jackknife+ baselines on a variety of biased, real-world datasets
Here, JAW maintains coverage at target level (90%) under covariate shift while jackknife+ does not👇🏽
Thoughts? -
Suchi Saria reposted
Nathan Ng
@learn_ng
We've got a great lineup of speakers including @YejinChoinka @bneyshabur @suchisaria @hhexiy @byron_c_wallace and @hendrycks!
If you're interested at all in helping to make sequential models generalize better stop by and chat! -
Suchi Saria reposted
Gianrico Farrugia
@GFarrugiaMD
Underrepresentation in #ai threatens the delivery of accurate, unbiased medical care. Mayo Clinic's @jhalamka and @plcerrato, joined by @drnigam and @suchisaria, share key solutions to more successfully develop fair and equitable algorithms: www.mayoclinicplatform.org/2022/11/15/the-algorithmically…
-
Suchi Saria
@suchisaria
ML friends at #twitter who had to face #twitterlayoffs @bayesianhealth is doing really exciting work in state-of-the-art multimodal AI & it’s use for improving healthcare & saving lives. Great culture; no loc reqs. DM or send CV to careers AT bayesianhealth.com if interested
-
Suchi Saria reposted
Daniel Gould, MD, PhD
@DJGould94
www.nature.com/articles/s41746-022-00597-7
Fascinating exploration of clinicians' views on their firsthand use of #ai decision aids. Interestingly, these clinicians did not see interpretability as being highly important as long as they had a functional understanding
@KatharineEHenry @suchisaria -
Suchi Saria reposted
Suhail
@Suhail
The AI spring has the most utility focused applications being built since mobile: new search engines, creativity tools, groundbreaking compression algos, 10x better code complete.
It feels like we finally broke into a new computing era. -
Suchi Saria reposted
Stanford CME
@StanfordCME
#stanfordaihealth is pleased to announce the 2022 Keynote Speakers!
On 12/6-12/7, we will be joined by experts Priscilla Chan of @ChanZuckerberg, Dr. John D. Halamka of @MayoClinic & Suchi Saria of @bayesianhealth
Register by 10/31 for a discounted rate: aihealth2022.stanford.edu -
Suchi Saria reposted
Vineeta Agarwala
@vintweeta
To all who celebrate, #happydiwali!
-
Suchi Saria reposted
Michael Calderwood, MD, MPH
@CalderwoodMD
Machine learning for identifying patients at risk of severe sepsis. Talk by @suchisaria at #idweek2022. Input + Output -> Rules. pubmed.ncbi.nlm.nih.gov/32166234/
-
Suchi Saria
@suchisaria
Haven’t been to a room this big in person since pre-COVID! About to give a talk here in 45 mins. Wish me luck! #idweek2022
-
Suchi Saria
@suchisaria
Very thoughtful discussion on our work on AI at the point of care in @TheAtlantic
www.theatlantic.com/health/archive/2022/10/sepsis-artific…
Thanks Simar Bajaj! -
Suchi Saria reposted
Aaron Levie
@levie
AI is clearly the future of software. Now the race is between existing software plugging in AI to make it smarter and new software designed from the ground up with AI as the differentiator.
-
Suchi Saria reposted
Johns Hopkins Medicine
@HopkinsMedicine
Doctors and researchers developed a system to show clinicians when someone is at risk for sepsis and suggests treatment protocols. r.socialstudio.radian6.com/8b9caa23-da42-47bd-91b9-532c1d…
-
Suchi Saria reposted
Simar Bajaj
@SimarSBajaj
I wrote a piece for @undarkmag on an #ai algorithm for sepsis—the leading cause of death worldwide
There are lots of machine learning tools, but not much real-world evidence. So @suchisaria's research at @JohnsHopkins is potentially exciting. Have a read
undark.org/2022/10/12/a-new-transparent-ai-tool-may-help-… -
Suchi Saria
@suchisaria
Giving the keynote at @Nature’s annual meeting tomorrow. It’s accessible virtually. Join!
Link to register: conferences.nature.com/event/0b87c6fb-30cf-495a-9da0-88b2… -
Suchi Saria
@suchisaria
🔥 FDA just released their latest guideline on oversight of decision support applications: www.fda.gov/media/109618/download
‼️ Their proposal seems to state that many EMR-embedded predictive tools (eg, solutions for sepsis and EOL) should be regulated.
🤔This has massive implications... -
Suchi Saria
@suchisaria
A PhD student right after they finish finish their PhD
Suchi Saria
@suchisaria
And assistant professors after they receive news of tenure!
-
Suchi Saria reposted
Nature Medicine
@NatureMedicine
Thinking of registering for our #digitalhealth conference #medinvirtualage22? (see link in comments)
@suchisaria will join as one of our keynote speakers; you may know her from these recent articles (which inspired our July cover):
www.nature.com/articles/s41591-022-01894-0
www.nature.com/articles/s41591-022-01895-z -
Suchi Saria reposted
Gabriele Corno
@Gabriele_Corno
Hugs are the universal medicine
-
Suchi Saria reposted
Buitengebieden
@buitengebieden
Massage time.. 😅
🎥 IG: menma_shiba -
Suchi Saria reposted
Matt Bishop
@MatthewTBishop
"When people think about scalability they think about productization of tech. They don't think of productization of the implementation experience and the productization of behavioral change." @suchisaria #onctechforum @ONC_HealthIT
-
Suchi Saria reposted
Matt Bishop
@MatthewTBishop
I'm stoked to see @suchisaria present @ONC_HealthIT #onctechforum! It's been 3 years since we met at a Data Science Healthcare informal meetup on a rooftop bar in Manhattan after meeting on Twitter. Sure you will crush it @suchisaria. @Bayesianhealth is doing outstanding work.
-
Suchi Saria reposted
Denis Wirtz
@deniswirtz
An applicant whose family has ties to Hopkins is not provided any advantage, i.e. there is no "legacy" students at @JohnsHopkins.
Shamefully, this is not the case for the vast majority of US universities and colleges, including many public universities.
Denis Wirtz
@deniswirtz
. @JohnsHopkins is ranked #7 in the latest USNWR ranking of US universities.
Hopkins has the highest rating for student excellence.
39% of incoming students are URM and 31% are first-generation, low-income -
Suchi Saria reposted
Mount Holyoke College President
@MHC_President
So proud of our alum Suchi Saria’04 @suchisaria saving lives with breakthrough algorithm designed to spot life-threatening sepsis infections before it’s too late #aisaveslives @mtholyoke
Johns Hopkins University
@JohnsHopkins
Sepsis contributes to one in three in-hospital deaths in the U.S.
A new algorithm designed at @JohnsHopkins can cut that number by 20%:
www.scientificamerican.com/article/algorithm-that-detects… -
Suchi Saria reposted
Bianca Caban
@bianca_caban
It was great getting together with our team, founders, and LPs to celebrate the launch of @aixventureshq. Shoutout to @shaunbjohnson and @fangy23 for putting together an inspiring and engaging agenda!
#ai #community #venturecapital -
Suchi Saria reposted
Dr. Faisal Rahman
@frahman480
An excellent review for #cardiologists in #machinelearning by ⭐ resident @aamirjavaidmd @hopkinsheart @CiccaroneCenter @JohnsHopkinsDOM @suchisaria @DoctorMarvelMD @rblument1 @AJPCardio doi.org/10.1016/j.ajpc.2022.100379
-
Suchi Saria reposted
AIX Ventures
@aixventureshq
A thought provoking keynote by @erikbryn, with intros from co-founders @RichardSocher & Shaun Johnson, about the economic implications of AI automation at our AIX Ventures Launch Event earlier this month.
We hope you enjoy it and leave comments below!
www.youtube.com/watch?v=w4zIZL5QiZk -
Suchi Saria reposted
Joao Monteiro
@JMinImmunoland
Our Nature Conference on Medicine in a Virtual Age is coming up, October 4-6, feat. keynotes talks by @suchisaria @doctorsoumya & @vivianleemd, discussing how the #pandemic accelerated transition of care from in-person to virtual & what future looks like. go.nature.com/3bHSZCS
-
Suchi Saria reposted
Denis Wirtz
@deniswirtz
Amazing story by the great @suchisaria and @withyouDrWu and their fellows
Johns Hopkins University
@JohnsHopkins
Sepsis contributes to one in three in-hospital deaths in the U.S.
A new algorithm designed at @JohnsHopkins can cut that number by 20%:
www.scientificamerican.com/article/algorithm-that-detects… -
Suchi Saria reposted
Johns Hopkins University
@JohnsHopkins
Sepsis contributes to one in three in-hospital deaths in the U.S.
A new algorithm designed at @JohnsHopkins can cut that number by 20%:
www.scientificamerican.com/article/algorithm-that-detects… -
Suchi Saria reposted
Patrick Malone, MD PhD
@patricksmalone
Designing an optimal process and interface for integrating AI and clinician is tricky. A trio of landmark papers published last week by @suchisaria's lab and @bayesianhealth provide some of the most impressive data yet on this issue.
Bayesian Health
@bayesianhealth
The opportunity for AI to improve health outcomes is massive. But what’s been missing is rigorous peer-reviewed studies showing AI used in practice. Today we published 3 studies in Nature Medicine that show for the first time ever how AI saves lives.
www.bayesianhealth.com/sepsis-research/ -
Suchi Saria
@suchisaria
So incredibly proud.
Bayesian Health
@bayesianhealth
The opportunity for AI to improve health outcomes is massive. But what’s been missing is rigorous peer-reviewed studies showing AI used in practice. Today we published 3 studies in Nature Medicine that show for the first time ever how AI saves lives.
www.bayesianhealth.com/sepsis-research/
Suchi Saria
@suchisaria
The key results, press releases and backstory here: www.linkedin.com/posts/suchisaria_aiforgood-intelligentca…
(I had planned to write a dedicated Twitter thread but overwhelmed by the press and inbound response.)
Suchi Saria
@suchisaria
This podcast does a great job of talking through the results with perspective from two wonderful physicians:
-
Suchi Saria reposted
Bayesian Health
@bayesianhealth
The opportunity for AI to improve health outcomes is massive. But what’s been missing is rigorous peer-reviewed studies showing AI used in practice. Today we published 3 studies in Nature Medicine that show for the first time ever how AI saves lives.
www.bayesianhealth.com/sepsis-research/ -
Suchi Saria reposted
Eric Topol
@EricTopol
There aren't many prospective trials of #ai in medicine. This week a new one for sepsis @NatureMedicine shows considerable promise for⬇️mortality
www.nature.com/articles/s41591-022-01894-0
www.nature.com/articles/s41591-022-01878-0
@suchisaria @withyouDrWu @KatharineEHenr @DBatesSafety @royjamesadams @AniaSyrowatka -
Suchi Saria reposted
Mari Devereaux
@maridevereaux
The longer a patient with sepsis waits for treatment, the more deadly their condition becomes. A new study suggests AI could help providers identify and treat sepsis sooner.
Modern Healthcare
@modrnhealthcr
#new: Data from @JohnsHopkins indicates that #artificialintelligence and #machinelearning are effective in alerting doctors earlier to signs of #sepsis in patients. h/t to our Safety & Quality Reporter @maridevereaux.
www.modernhealthcare.com/safety-quality/ai-can-reduce-sep… -
Suchi Saria reposted
Nature Medicine
@NatureMedicine
The prospective evaluation of a #machinelearning based early warning system for #sepsis showed that interaction of health care providers with the system was associated with better patient outcomes, including reduced in-hospital mortality @suchisaria www.nature.com/articles/s41591-022-01894-0
-
Suchi Saria reposted
JHTV
@JHTechVentures
"In 82% of #sepsis cases, the AI was accurate nearly 40% of the time. Previous attempts to use electronic tools caught less than half that many cases and were accurate 2% to 5% of the time." Exciting news from @bayesianhealth and @suchisaria! r.socialstudio.radian6.com/0c5ad5a9-ee1d-489c-bcc6-6313e5…
-
Suchi Saria reposted
npj Digital Medicine
@npjDigitalMed
Clinicians perceive #machinelearning as an augmentation rather than substitute to their clinical practice. This qualitative analysis by @KatharineEHenry, @suchisaria et al. explored opinions of clinicians working with a ML-based sepsis warning system...
www.nature.com/articles/s41746-022-00597-7 -
Suchi Saria reposted
AIX Ventures
@aixventureshq
Today we're thrilled to announce @suchisaria has joined AIX Ventures as its newest Investing Partner. Check out @RichardSocher's announcement for more info.
Welcome, Suchi!
www.aixventures.com/announcements/welcoming-our-newest-in… -
Suchi Saria reposted
Meanwhile in Canada
@MeanwhileinCana
A Blue Whale was spotted off the coast of Vancouver Island yesterday. The largest animal to ever grace this planet. The largest dinosaur weighed around 85 tons. Some male Blue Whales weigh as much as 200 tons. A Blue Whale's heart is 6 feet wide and weighs 400 lbs.
-
Suchi Saria reposted
Limor Gultchin
@gultchin
Final reminder: ~24 hrs left to submit your work (end of day May 27th, AoE)!
@aahladpuli @wald_yoav @victorveitch @Maggiemakar @angelamczhou @marikgoldstein @ShalitUri @suchisaria #causaltwitter #causalinference #machinelearning #icml2022 @slashML @trustworthy_ml
Limor Gultchin
@gultchin
Now that we're post-NeurIPS...
Consider submitting work to the ICML 2022 workshop SCIS: Spurious Correlations, Invariance and Stability!
We extended the submission deadline to May 27th, all details here: sites.google.com/view/scis-workshop/home -
Suchi Saria reposted
John Collison
@collision
As you become an adult, you realize that things around you weren't just always there; people made them happen. But only recently have I started to internalize how much tenacity *everything* requires. That hotel, that park, that railway. The world is a museum of passion projects.
-
Suchi Saria
@suchisaria
Congrats @mike_schatz!
Mike’s my 4th close colleague/friend on Time 100! If you want to get on the Time 100, you know what you need to do! 😉 😎 Obviously, I mean do fabulous work…
Denis Wirtz
@deniswirtz
@JHU_BDPs Michael Schatz among 'TIME 100' most influential people in the world | Hub hub.jhu.edu/2022/05/23/michael-schatz-time-100-influentia…
-
Suchi Saria
@suchisaria
I was watching the last season of @SHO_Billions and heard the “I trained under MacArthur genius @DaphneKoller!” Warmed my ❤️ that they did their homework.
-
Suchi Saria reposted
Danilo Bzdok
@danilobzdok
DECIDE-AI: I was happy to contribute to #reporting #guidelines for decision support systems driven by #ai
www.nature.com/articles/s41591-022-01772-9
www.bmj.com/content/377/bmj-2022-070904.full?ijkey=PUU21u…
@suchisaria @MaartenvSmeden @GSCollins
@NatureMedicine @JAMA_current -
Suchi Saria reposted
Dr. Faisal Rahman
@frahman480
Our analysis presenting a novel algorithm using #ehr data to predict cardiogenic #shock early - published in @MyJSCAI. Need to treat #scai < C shock early to prev deterioration & #mortality authors.elsevier.com/sd/article/S2772-9303(22)00292-7
@hopkinsheart @ngilotraMD @suchisaria @BCMHeart #cardiotwitter -
Suchi Saria reposted
AppliedMicrobiologyResearch
@AppliedMicrobi2
Chairing the keynote lecture @ECCMID #eccmid2022 focusing on #ai and #medicine by @suchisaria - fantastic and insightful talk. Thank you!
Time to teach about machine learning, its #possibilities and #challenges! -
Suchi Saria
@suchisaria
About to do a live video interview. It’s for a massive clinical meeting (15k+ attendees). Lost track of time and got out of bed 5 mins prior to the interview! I’m here - before the interviewer is ready for me. Phew!!!!
-
Suchi Saria
@suchisaria
Found this so inspiring! Enjoy! www.linkedin.com/posts/suchisaria_machinelearning-automat…
-
Suchi Saria reposted
NEJM Catalyst
@nejmcatalyst
.@suchisaria @bayesianhealth introduces session 2: AI for Augmenting Frontline Providers #nejmcatlive
Read her paper here: catalyst.nejm.org/doi/full/10.1056/CAT.22.0075 -
Suchi Saria reposted
Jayson Marwaha, MD MSc
@Jayson_Marwaha
Tomorrow (Mar 24), we’ve been invited to share our work on translating ML tools to the bedside at the
@NEJMcatalyst/@AdvisoryBd event on #ai for Healthcare Delivery. Our associated paper here: catalyst.nejm.org/doi/full/10.1056/CAT.21.0477
Feat. amazing speakers @arieldora @willjg @suchisaria et al! -
Suchi Saria reposted
NEJM Catalyst
@nejmcatalyst
Emerging best practices for the successful adoption and use of #artificialintelligence in health care, and a key missing ingredient: rigorous evaluation: catalyst.nejm.org/doi/full/10.1056/CAT.22.0075 #healthcareai via @suchisaria @JohnsHopkins @bayesianhealth @JHUMCEH @JHUCompSci @HopkinsEngineer
-
Suchi Saria
@suchisaria
I really enjoyed doing this interview. If you’re an AI researcher, user, or developer, it may be interesting to learn about moving research from a paper to a working product in a complex industry like healthcare.
Day Zero Podcast
@DayZeroPodcast
Dr. Michael D. Abramoff shares his 5 year outlook for autonomous AI in healthcare.
On this weeks #dayzero episode, hosted by @suchisaria, Dr. Abramoff shares how his company Digital Diagnostics developed the first FDA approved autonomous AI
Listen here - toto22up.com/day-zero/ep-24-ethical-autonomous-ai-diagnos… -
Suchi Saria reposted
Day Zero Podcast
@DayZeroPodcast
#meetthehost of tomorrow's #dayzero episode - @suchisaria #founder and #ceo of @bayesianhealth and the Director of the Machine Learning and Healthcare Lab at @JohnsHopkins
Learn more about Dr. Saria here - toto22up.com/day-zero/ep-7-meet-the-hosts-suchi-saria-ph-d/ -
Suchi Saria
@suchisaria
This is going to be terrific. Join us!
NEJM Catalyst
@nejmcatalyst
Our March 24 #healthcareai event with @advisorybd is soon! Looking forward to hearing from @bratogram @ronlivs @suchisaria @arieldora @christinasilcox and more on enhancing public health, augmenting the front line, and adoption barriers: events.nejm.org?promo=OCFK322H&query=cat-social #nejmcatlive
-
Suchi Saria
@suchisaria
Another reason why black box models built without domain understanding and without high quality ongoing monitoring are not the answer! AI in health requires the same level of caution and intentional development as we put into developing and disseminating medicines.
-
Suchi Saria
@suchisaria
Want to use your #ml skills to build data products that *actually* save lives? We're building a superb team over @bayesianhealth & are looking to add 2 ML data product designers/engineers (with >2+ years of industry experience) immediately. More here: www.bayesianhealth.com/careers/
Bayesian Health
@bayesianhealth
3 more hours to heading out for our Bayesian beach retreat at the #isleofpalms South Carolina! Lots to celebrate! 🎉
Suchi Saria
@suchisaria
We're built to be a virtual-first company (multiple yearly on-sites at fun locations — pics from last on-site location☝️, weekly games, instrumented for distributed communication). This allows for a work hard + play hard culture in a pretty cool way!
-
Suchi Saria
@suchisaria
Hahahahahahahahahahahaha
Ferenc Huszár
@fhuszar
learning theory vs deep learning
-
Suchi Saria reposted
Paul Graham
@paulg
If you're a student interested in both technology and startups and wondering how to combine them, the answer is: serially. Focus entirely on technology at first. Then once you're an expert in some technology (which will take several years), you can start to think about startups.
-
Suchi Saria
@suchisaria
👇at least in healthcare, when we look back, the rapid adoption of AI is going to feel very similar. It’s so desperately needed to scale up (high quality) care delivery…
-
Suchi Saria
@suchisaria
Most things that are free — especially as it relates to data — come with a catch-22 because there needs to be some model for sustaining the investment. I don’t know if there is one here (@oziadias ?) but otherwise this is a much needed, fantastic effort by terrific researchers.
-
Suchi Saria reposted
Nikhil Krishnan
@nikillinit
New OOP featured job posting:
@bayesianhealth is building AI products for health systems to give physicians, nurses, and care team members real-time clinical signals and alerts. They're hiring for product in all levels from Senior to VP of Product
out-of-pocket-jobs.pallet.xyz/jobs/0fe35dc2-bce9-438c-8a1… -
Suchi Saria reposted
Vala Afshar
@ValaAfshar
In 1995, only two years after the birth of the worldwide web, @BillGates tried to explain ‘The Internet’ to Letterman
Technology pioneers today face the same difficulty and ridicule as they describe AI and blockchain -
Suchi Saria reposted
Denis Wirtz
@deniswirtz
Today, we announce a major gift from Mike Bloomberg, which will facilitate the cluster recruitment of 2 named professors and 2 junior faculty in Artificial Intelligence and Society at @JohnsHopkins.
Cluster co-led by Pr.Anton Dahbura, Lingxin Hao, Cara LaPointe and Debra Mathews -
Suchi Saria
@suchisaria
Tackling dataset shift is one of the most important topics for moving AI into real world settings.
I’ll give an overview of our 6 yrs of work here going from theory to practice to policy, and where I see the the open frontiers are, working backwards from health policy
Pang Wei Koh
@PangWeiKoh
Join us at the NeurIPS Workshop on Distribution Shifts (DistShift) next Monday, featuring 7 invited talks, a panel on future directions, and 85 accepted papers!
When: Dec 13, 9am-6pm PT
Website: sites.google.com/view/distshift2021
Virtual site (needs registration): neurips.cc/virtual/2021/workshop/21859 -
Suchi Saria reposted
Bayesian Health
@bayesianhealth
So thrilled that @Rock_Health selected our founder and CEO Suchi Saria to their list of Top 50 in Digital Health as a builder building the future of digital health. We were selected because of our outcomes and our rigorous research first methodology.
www.top50indigitalhealth.com/2021honorees -
Suchi Saria reposted
Julie Yoo
@julesyoo
Congrats to the @a16z bio+health peeps in 2021 @Rock_Health Top 50: @SavageLucia (@OmadaHealth), Varun Ganapathi (@AKASAHealth), @SuchiSaria (@BayesianHealth), @ImanAbuzeid (@JoinIncredible) - www.top50indigitalhealth.com/2021honorees
-
Suchi Saria reposted
Day Zero Podcast
@DayZeroPodcast
Meet the Advisory Council! Our hosts this season will include @rich_roth, Lynne Chou O’Keefe, Marcus Osborne, Don Trigg, Nathan Bays, @aaronmartinfc, @suchisaria & @julesyoo.
Our first host of the season is Julie Yoo! Learn more about her here: toto22up.com/day-zero/ep-1-meet-the-hosts-julie-yoo/ -
Suchi Saria reposted
Oded Rechavi
@OdedRechavi
The post doc with the Cell paper arriving at his lab meeting
-
Suchi Saria
@suchisaria
👇especially practitioners / health system leaders
Karandeep Singh
@kdpsinghlab
What are you doing on Nov 18th?
Hopefully attending SMILE, a symposium dedicated to ML implementation & evaluation!
Featuring:
@drnigam
@ctlin1
Anna Goldenberg, PhD
@shalmali_joshi_
@MarkSendak
@AnthonyChangMD
@rkamaleswaran
and more!
Register here: www.eventbrite.com/e/symposium-for-machine-learning-imple… -
Suchi Saria
@suchisaria
Two of my favorite/most exciting developments in software:
#1. software that saves lives (aka what @bayesianhealth does) — easy to see why.
#2. VR spaces (eg meta) — it has the potential to change the way we experience life including travel, work, exercise, & socialize.
Aleks Lyng
@alekslyng
I just watched the entire meta keynote by Zuck. It was incredible. Huge props to him for ignoring the noise and not being afraid to disrupt himself.
Don’t get recency bias. Zuck is on the Mt Rushmore of Founder CEOs.
I may start buying $FB again. -
Suchi Saria reposted
Matt Sakumoto
@MattSakumoto
Great session on "Making Health AI Work in the Real World" Impressive expert panel including @oziadias @EricTopol @suchisaria @kdpsinghlab @pyhsueh #amia2021
-
Suchi Saria reposted
cristina brodu
@CriB3001
AI Applications:
@A_Aspuru_Guzik from self driving lab for optimization of materials and drug discovery, to robots performing chemical reactions in future
@suchisaria data from 100s of integrated sensors on the body can permit diseases early detection, be a clinical decision tool
Harvard Radcliffe Institute
@RadInstitute
From science and engineering to art, how can AI be used? Join @A_Aspuru_Guzik, @suchisaria, @VanArman, and @tsimonite as they discuss the possibilities in “AI Applications.”
-
Suchi Saria reposted
Eric Musser
@EricJMusser
Great session at @HLTHEVENT on #ai’s role in bias and equity. Going to be an inspiring few days! @suchisaria @TCLawry @jkvedar #hlth2021
-
Suchi Saria reposted
Ben Tappin
@Ben_Tappin
Happy to share a new preprint with @AdamBerinsky & @DG_Rand.
We find that exposure to arguments and evidence (1) persuades U.S. partisans and (2) remains similarly persuasive despite countervailing cues from in-party leaders (Trump, Biden).
🧵 1/12
psyarxiv.com/247bs/ -
Suchi Saria reposted
Vaishali
@vmittal26
This is sad.
Ritu Thamman MD
@iamritu
💔women wait 10 min longer to be seen for chest pain in ER
💔women less likely to be triaged as emergent case
💔women less likely to get ekg,or🫀monitor, or to
be seen by a cardiologist in the ER for CP
💔less likely to be admitted 🏥 for CP #accwic #acc21
#disparitiesofcare -
Suchi Saria reposted
Bayesian Health
@bayesianhealth
.@suchisaria talked with @sethbjoseph about #ai in #healthcare progress; we're now at a time where research-backed AI solutions that are rigorously evaluated can improve care delivery and improve patient outcomes in a sustainable way. www.forbes.com/sites/sethjoseph/2021/09/30/artificial-int…
-
Suchi Saria
@suchisaria
Having a lovely Saturday:
Brain is racing … it’s quiet and sunny outside!
My two favorite things. -
Suchi Saria
@suchisaria
When @_asubbaswamy joins the ranks faculty or begins to mentor students, his mentees will have someone amazing to learn from.
As his PhD advisor, I’ve loved uncovering great nuggets around research or advising from him! #academictwitter
Adarsh Subbaswamy
@_asubbaswamy
Overall agree with the thread's point that grad students need to be willing to repeatedly take feedback and improve their work. But I think the focus on "work is always unfinished" really misses the mark, especially in the context of higher education. 1/7
-
Suchi Saria
@suchisaria
Fascinating! Was wondering how someone came to realize ways to measure how far a star is.
en.m.wikipedia.org/wiki/Henrietta_Swan_Leavitt -
Suchi Saria
@suchisaria
I lost 10 Twitter followers when I declared I was going on vacation. I’ve clearly curated a type.
Suchi Saria
@suchisaria
Going on vacation for a few days after a VERY long time. Excited!!!
-
Suchi Saria
@suchisaria
Going on vacation for a few days after a VERY long time. Excited!!!
-
Suchi Saria reposted
Bob Wachter
@Bob_Wachter
I still believe we're entering an era of digital transformation in healthcare. 2 years ago, I would have given EHR vendors (Epic, Cerner...), Digital Giants (Google, Apple, MSFT...), & VC-funded start-ups equal odds of success. Now, I'd give start-ups the edge. Focus is key.(3/3)
-
Suchi Saria reposted
Bayesian Health
@bayesianhealth
Incredibly excited that @BusinessInsider named @suchisaria to their 30 under 40 in Healthcare list! Our mission is big - but together we're going to change patient care and save lives. #healthtech #digitalhealth www.businessinsider.com/30-leaders-under-40-changing-heal…
-
Suchi Saria
@suchisaria
Very honored and thrilled to be listed among these other really inspiring leaders.
Healthcare is hard. Health-tech is even harder. But, I LOVE the work we're doing every day to save lives with @bayesianhealth!
Bayesian Health
@bayesianhealth
Incredibly excited that @BusinessInsider named @suchisaria to their 30 under 40 in Healthcare list! Our mission is big - but together we're going to change patient care and save lives. #healthtech #digitalhealth www.businessinsider.com/30-leaders-under-40-changing-heal…
-
Suchi Saria reposted
Dimitri Dadiomov
@dadiomov
Microsoft was 20 years old when Windows 95 shipped.
Intel was 25 years old when Pentium came out.
Apple was 31 years old when iPhone launched.
All were still run by their founders.
There’s SO much innovation still to come from today’s 10-15 year old, founder-run companies. -
Suchi Saria
@suchisaria
Twins in my family, one got her surgery @MDAndersonNews, the other at a leading Boston AMC (name removed).
At dinner, family describing their experience w/ each —
at the Boston AMC, they felt they were “thrown to the wolves” Yikes…
at MD Anderson, “they took such good care”!! -
Suchi Saria
@suchisaria
Thread 👇 on health + AI, evaluation mistakes, current state (TLDR; not great), and ideas for path forward.
Great reporting by @statnews @caseymross with thoughtful insights from several experts.
@PencinaPhd “Our naivete has enabled this kind of behavior.”
Bayesian Health
@bayesianhealth
Building & deploying predictive tools in healthcare isn’t easy. The data is challenging from the start and building models that can analyze this data requires a deep understanding of the latest AI/ML strategies & an ability to employ it effectively. (1/4)
www.statnews.com/2021/07/26/epic-hospital-algorithms-seps… -
Suchi Saria reposted
Bayesian Health
@bayesianhealth
And because it's us, we had to have some sort of competition - last night, it was an Iron Chef cooking competition with a secret ingredient! Can you guess what it was⁉️ #bayesianbeachretreat 🍽️👩🍳🏖️
-
Suchi Saria reposted
Bayesian Health
@bayesianhealth
It's been amazing spending time together as a team at our Bayesian Beach Retreat! #jumpingforjoy 🏖️
-
Suchi Saria reposted
The TWIML AI Podcast
@twimlai
Today we’re joined by @SuchiSaria of @BayesianHealth and @JHUMCEH @JHUCompSci to discuss her work at the intersection of ML and healthcare, the holdup on broad ML adoption in healthcare, and of course, the work she’s doing at Bayesian Health. twimlai.com/ml-innovation-in-healthcare-with-suchi-saria
-
Suchi Saria reposted
Nika Haghtalab
@nhaghtal
Excited to attend the ICML Debate, Tuesday @ 9am PT.
With Charles Isbell(@isbellHFh), Ti John, Michael Kearns(@mkearnsupenn), Shakir Mohamed(@shakir_za), Stephen Roberts, Suchi Saria (@suchisaria), Rich Sutton, and Martha White.
icml.cc/virtual/2021/social/12458
#icml2021 #aiethics -
Suchi Saria reposted
-
Suchi Saria reposted
Eric Topol
@EricTopol
—Explainability drawbacks
science.sciencemag.org/content/373/6552/284
@ScienceMagazine @gerke_sara @CohenProf
—Dataset shifts
www.nejm.org/doi/full/10.1056/NEJMc2104626?query=featured… by @IAmSamFin
@zakkohane @kdpsinghlab @suchisaria @NEJM
Summarized so well by @DrXiaoLiu's 🧵
x.com/DrXiaoLiu/status/1416770766859821057 /2
Xiao Liu
@DrXiaoLiu
Earlier this week @IamSamFin @kdpsinghlab @zakkohane @suchisaria et al published a excellent letter in @NEJM: Checklist for recognising clinical dataset shift for #ai
This is a really important topic & I recommend properly combing through Suppl Table 2.
Some thoughts: 1/n
Sam Finlayson
@IAmSamFin
Letter: "The Clinician and Dataset Shift in Artificial Intelligence" out in @NEJM w/ @_asubbaswamy , @suchisaria, @kdpsinghlab, @zittrain, @zakkohane, et al
Link: www.nejm.org/doi/full/10.1056/NEJMc2104626
Full version of Table (w/ details+citations): sgfin.github.io/assets/dataset_shift/NEJM_DS_Supplement.pdf
Brief thread: 1/n -
Suchi Saria reposted
Xiao Liu
@DrXiaoLiu
Earlier this week @IamSamFin @kdpsinghlab @zakkohane @suchisaria et al published a excellent letter in @NEJM: Checklist for recognising clinical dataset shift for #ai
This is a really important topic & I recommend properly combing through Suppl Table 2.
Some thoughts: 1/n
Sam Finlayson
@IAmSamFin
Letter: "The Clinician and Dataset Shift in Artificial Intelligence" out in @NEJM w/ @_asubbaswamy , @suchisaria, @kdpsinghlab, @zittrain, @zakkohane, et al
Link: www.nejm.org/doi/full/10.1056/NEJMc2104626
Full version of Table (w/ details+citations): sgfin.github.io/assets/dataset_shift/NEJM_DS_Supplement.pdf
Brief thread: 1/n -
Suchi Saria reposted
Jenn Wortman Vaughan
@jennwvaughan
I've seen several tweets this week suggesting that newcomers stay away from ML research unless they already have a killer idea.
Please don't take this advice.
Almost no one starts PhD research with a clear idea of what they'll do, and those who do often change directions. (1/3) -
Suchi Saria reposted
Sam Finlayson
@IAmSamFin
Letter: "The Clinician and Dataset Shift in Artificial Intelligence" out in @NEJM w/ @_asubbaswamy , @suchisaria, @kdpsinghlab, @zittrain, @zakkohane, et al
Link: www.nejm.org/doi/full/10.1056/NEJMc2104626
Full version of Table (w/ details+citations): sgfin.github.io/assets/dataset_shift/NEJM_DS_Supplement.pdf
Brief thread: 1/n -
Suchi Saria reposted
MedCity News
@medcitynews
A lot of #ai tools are finding a place in healthcare, so how can a company stand out? @JohnsHopkins spinout has data on how its tool is used by physicians, and the #startup is now out of stealth, @elisereuter reports medcitynews.com/2021/07/johns-hopkins-spinoff-looking-to-… #artificialintelligence #healthtech
-
Suchi Saria reposted
Jay Rughani
@JayRughani
My parents lost a son due to complications related to sepsis on Sep 12, 1985 before I was born. Sadly, too many families have stories like this.
So I'm particularly inspired by the results of @bayesianhealth's prospective study on their early warning system for sepsis: -
Suchi Saria reposted
DavisLiuMD 劉徳偉🇨🇦 🇹🇼 🇺🇸
@davisliumd
💯 key leadership skill: being comfortable with debate and being uncomfortable! @suchisaria
When things aren’t working, when people don’t listen, I don’t start from a place of feeling dejected or surprised, I start from a place of, “That’s expected. Now, what am I going to do?”
Julie Yoo
@julesyoo
3/ @BayesianHealth's focus on building physician trust was clear in the study results, which showed 89% user engagement. great coverage in @statnews as well: www.statnews.com/2021/07/12/bayesian-health-artificial-in…
-
Suchi Saria reposted
Vineeta Agarwala
@vintweeta
"Founder and CEO @suchisaria...is setting a new standard for the level of clinical rigor that digital health products must exhibit to pass muster in high-stakes, patient-facing clinical settings." 👏👏
Congrats @bayesianhealth on your launch; excited to be working together!
Vijay Pande
@vijaypande
We’re now at a tipping point where AI can supercharge clinician decision making. @bayesianhealth has brought this new capability to the front lines of healthcare with demonstrable results. Thrilled to lead Bayesian's Series A & partner with @suchisaria !
a16z.com/2021/07/12/investing-in-bayesian-health/ -
Suchi Saria reposted
Vijay Pande
@vijaypande
We’re now at a tipping point where AI can supercharge clinician decision making. @bayesianhealth has brought this new capability to the front lines of healthcare with demonstrable results. Thrilled to lead Bayesian's Series A & partner with @suchisaria !
a16z.com/2021/07/12/investing-in-bayesian-health/ -
Suchi Saria
@suchisaria
I’ve often heard the recommendation of banning sensitive variables like race as algorithmic inputs.
But, statistically speaking, not including race as a variable can lead to algorithms that are “racist”. So banning inclusion of sensitive variables doesn’t seem right.
Thoughts? -
Suchi Saria
@suchisaria
Excited to give the keynote lecture this afternoon @ ISBA (International Society for Bayesian Analysis) world meeting: events.stat.uconn.edu/ISBA2021/ My talk is on "Making Health AI work in the Real World".
In 2011, I gave my 1st talk there as a fresh grad & remember being v. nervous! -
Suchi Saria
@suchisaria
Very nice work by @kdpsinghlab & team on evaluating Epic's widely deployed sepsis tool. See thread below on issues to consider when evaluating offline model performance and also common ways in which previous studies have erred.
Karandeep Singh
@kdpsinghlab
Thank you for folks who have shared or commented on our paper. I know the paper is being used by some to dunk on Epic. Rather than piling on, I want to provide a clear-eyed view of what we found, what it means, and what I would suggest to Epic (& other model devs) going forward.
Post not available here, view on X
Suchi Saria
@suchisaria
If performance is indeed bad & if so many systems are using the tool, why wasn't this caught before?
In my conversations w/ system leaders using Epic's tool, they've selected high thresholds to reduce false alerts. But this means they're operating at very low sensitivity (<12%)
Suchi Saria
@suchisaria
And it is hard to measure sensitivity using qualitative evaluations. You need a good labeler with offline or online monitoring to measure it. As a result, qualitative evaluations can't account for lowered efficacy b/c of low sensitivity.
Post not available here, view on X
Suchi Saria
@suchisaria
Should studies like this be dismissed b/c they're single center? Yes, we should be more skeptical if there were other prior rigorous, large studies contradicting this finding.
What's been surprising is that today tools are scaling without evidence.
Karandeep Singh
@kdpsinghlab
Our biggest limitation is that our results come from a single center.
Tho I know of another center with a similar AUC to us using similar code, we did not combine into 1 paper bc their scoring started shortly before COVID, which introduces some complications. More to come...
Suchi Saria
@suchisaria
High quality digital tools will be game changing for improving outcomes. But, rigor in measurement and high quality evidence is key to gaining provider trust and adoption.
(COI: I am an equity holder in several digital health companies including one that works on sepsis.) -
Suchi Saria reposted
Gaurav Singal, MD
@GauravSingalMD
Excited to debate utility of rwd and how health systems should engage with @vintweeta, @vijaypande, @JorgeCondeBio, @julesyoo, @atulbutte, @suchisaria, & @truveta’s Terry Meyerson tmw, Jun 21 at 8:30 PM EDT on @clubhouse. Join us! www.clubhouse.com/event/MKjGNvvQ
-
Suchi Saria reposted
Dean Sittig
@deansittig
We need much more evaluation of these types of algorithms before they are adopted. #saferehrs @histalk @HardeepSinghMD @VUMC_VCLIC @suchisaria www.fastcompany.com/90641343/epic-deterioration-index-alg…
-
Suchi Saria
@suchisaria
Generally, my comments are in AI or health but this time I talk about how startups are evolving post pandemic.
Since this interview, we’ve already upgraded our upcoming retreat to a giant beach house in South Carolina this summer! ❤️ @bayesianhealth
www.wsj.com/articles/tech-startups-ditch-the-office-for-f… -
Suchi Saria reposted
Karandeep Singh
@kdpsinghlab
Come join us and reflect on the ways (good and bad) in which AI is inserting itself into healthcare.
Should be a great discussion!
marketing.statnews.com/artificial-intelligence-transformi… -
Suchi Saria
@suchisaria
Congratulations to you both and so excited for Victor to be joining us!
Ehsan Hoque
@ehsan_hoque
What a joy it has been to work with @VictorNAntony, one of the most selfless and committed team members that I have ever worked with. Victor is off to @JohnsHopkins to work with @chienming_huang and @SuchiSaria (2/4)
-
Suchi Saria reposted
Isaac Kohane
@zakkohane
Where AI meets clinical usefulness: SAIL sail.health/index.html symposium October 2021 in Bermuda: Call for Abstracts sail.health/event/sail-2027/participate/. Accepted abstract is ticket of entry to meeting of clinical implementers, clinical publishers (e.g. @NEJM @TheLancet) & AI researchers
-
Suchi Saria
@suchisaria
One more opportunity—a clinician scientist w/
- experience writing high quality papers
- wants to partner with an amazing interdisciplinary team to turn analyses into actionable insights
The position allows clinicians to continue their practice & have other engagements.
DM 👌🏾
Suchi Saria
@suchisaria
Our stellar team of scientists, product & domain experts @bayesianhealth are seeking to grow & looking for 1) data scientists (with strong data eng chops and clinical data curiosity) and 2) part/full time time UX/HCI. Ping me or careers@bayesianhealth.com if you’re interested
-
Suchi Saria
@suchisaria
At a quest diagnostics, came here for labs. I show them my prescription. They say — cannot work with me without a printout!
I ask if I can email? — no!
can I print? — no!
So I figure out eFax from my phone to their fax.
#wowhc we invented vaccines but have the worst process -
Suchi Saria
@suchisaria
How do you scale high quality decision making without being a top down team (like most academic teams tend to be)?
It’s a hard org puzzle & a squishy topic that I’ve developed deep appreciation for in the process of building @bayesianhealth
Fantastic thread.
Dan Rose
@DanRose999
What defines a great company culture? I worked for two iconic companies and founders with nearly polar opposite cultures. Amazon was heads-down, secretive, forthright. Facebook was open, transparent, collaborative. Here's what I learned about culture working for Bezos and Zuck:
Post not available here, view on X
Suchi Saria
@suchisaria
But ... team developing and scaling novel breakthrough research-led innovations requires:
- staying motivated through failures (in research this is norm, not exception)
- rigor & critical thinking
- forging ahead when there aren’t any obvious right answers 3/
Suchi Saria
@suchisaria
And it’s hardest when victories require multiple years of work & many iterations (a lot of drug discovery & health AI work is of this kind).
But it’s been thrilling to work through this puzzle.
Have you had to think through this and if so, what have you found to be effective? -
Suchi Saria
@suchisaria
Our stellar team of scientists, product & domain experts @bayesianhealth are seeking to grow & looking for 1) data scientists (with strong data eng chops and clinical data curiosity) and 2) part/full time time UX/HCI. Ping me or careers@bayesianhealth.com if you’re interested
Suchi Saria
@suchisaria
Ideal candidates are excited to work at a startup, continual learners, energized by solving hard problems as part of a highly collaborative multidisciplinary team, and ecstatic about improving human health & outcomes.
-
Suchi Saria
@suchisaria
Revised version, now out at @Aistats2020 and includes feedback from our FDA colleagues. Very excited about this work.
arxiv.org/abs/2010.15100
Adarsh Subbaswamy
@_asubbaswamy
New preprint w/ (co-first author) @royjamesadams and @suchisaria: "Evaluating Models Robustness Under Dataset Shift" arxiv.org/abs/2010.15100
How can we evaluate *ahead of time* whether or not a model's performance will generalize from training to deployment? 1/ -
Suchi Saria
@suchisaria
😂 this faithfully captures what last year has felt like!
x.com/buitengebieden_/status/1374445042367033354/video/1 -
Suchi Saria reposted
James Landay
@landay
AI and models need to change after deployment. There are lots of things that shift after deployment that can lead to biases. @suchisaria has some key papers here!
#augmenthai -
Suchi Saria
@suchisaria
I’ve run into this issue for years ...
I’ve chalked it up to (a) incentive structures encourage ownership over citations or (b) my inability to tell clear narratives that are memorable.
It’s disheartening .. & worse when you know they’ve read the papers & haven’t cited it.
Dan Roy
@roydanroy
I'm getting tired of attending talks by people working directly in my area and having them reexplain to me ideas in my papers, specialized to their situation, without any awareness. It's really easy to tell when they glossed over the result. 8 pages people!
Suchi Saria
@suchisaria
I don’t enjoy calling it out. It feels too self-serving. And everything’s probably compounded by the stat that people are much less likely to listen to minority/female and more likely to attribute ideas to majority counterparts. Calling that out also feels self-serving.
-
Suchi Saria
@suchisaria
I have to give a 5 min (!) spotlight talk at 9:30am. I can’t seem to fall asleep ... I’ve given 250+ actual talks in my life so the issue isn’t nervousness ... Revealing some new preliminary data so too excited for my own good.
-
Suchi Saria
@suchisaria
What are really solid reads on building enterprise co? Any topic-GTM, branding, culture, growth tricks, bio
cc @katgleason @dadiomov @rbhar90 @zachweinberg @levie @paulg @julesyoo
My recs
-Challenger Customer
-Monetizing Innovation
-Career Imprints
-Multipliers
-Crucial Convos -
Suchi Saria
@suchisaria
😂 😂 made my day!
Yet, we gladly partake...
Daniela Witten
@daniela_witten
I just received an invitation to write a book chapter, which mentioned that I will be "compensated" with a hard copy of the book.
Academia: I can't even -
Suchi Saria
@suchisaria
In speaking w/ Dr. @leorahorwitzmd learned an appalling fact!
Many COVID vaccine doses are being thrown away b/c patients who’ve recently received another vaccine can’t receive COVID vaccine & current appointment process doesn’t ask for patient’s recent vaccine status!
Wow!
Suchi Saria
@suchisaria
This should be an easy fix using health IT!
-
Suchi Saria
@suchisaria
This is an exciting NIH workshop on how personal nutrition is evolving from better measurement!
Experts (more exciting than me) on this topic will be giving you a sense for state of the art on all aspects of this fast evolving field. And, it's all online!
Bruce Y. Lee
@bruce_y_lee
This coming Monday & Tuesday will be the @NIDDKgov @nih_nhlbi @NIH Precision Nutrition that I’ll be co-chairing. Lots of great speakers including @suchisaria @JHUCompSci who will be talking about her impressive work in #machinelearning in #healthcare
www.niddk.nih.gov/news/meetings-workshops/2021/precision-… -
Suchi Saria reposted
Jorge Conde
@JorgeCondeBio
This is wild: 1 dose of a DNA base editor ‘cures’ “Benjamin Button” aging disease (in mice) by correcting the disease-causing mutation 💉 🧬 🐁
This mutation has been known since 2003, but we now have the tools to correct it (and similar errors)
This is programmable medicine
David R. Liu
@davidrliu
Today we report the use of base editing in patient-derived cells and in mice to correct the most common cause of progeria, the devastating rapid aging disease. Progeria is typically caused by a dominant negative C•G-to-T•A point mutation in LMNA. 1/11 drive.google.com/file/d/1oH4W0z-PtsEcds70QIq-3XQxKlWcnax2…
-
Suchi Saria reposted
Esther Choo, MD MPH
@choo_ek
One year ago, Wuhan experienced a mysterious cluster of cases of pneumonia.
A year later, we have identified & characterized the virus and developed two effective and safe vaccines and started to administer them.
We have to stop and appreciate this breathtaking accomplishment. -
Suchi Saria reposted
Vijay Pande
@vijaypande
Amazing to hear the Moderna vaccine story directly from @sbancel -- the story of the year. Thank you @JorgeCondeBio and @omnivorousread !
a16z
@a16z
Breaking: The day after the FDA advisory committee recommends approval, Moderna CEO @Sbancel tells the story of developing the Covid-19 vaccine + the tech that powered the process, to @JorgeCondeBio & @omnivorousread on a special episode of Bio Eats World: future.com/podcasts/moderna-covid-vaccine-mrna-technology/
-
Suchi Saria
@suchisaria
I'll be looking for postdoc fellows who are interested in advancing technical & practical work on reliability, robustness and actionability of ML decision support (in partnership with FDA and NSF). You'll be evaluating methods on clinical apps. DM me. Remote employment is fine.
Suchi Saria
@suchisaria
I love working with people who are warm, kind, and fun! And, I'm quite approachable but I love to debate.
-
Suchi Saria reposted
Alan Karthikesalingam
@alan_karthi
Another great paper by @_asubbaswamy @royjamesadams @suchisaria. In healthcare we expect model deployment to present challenges not seen during model development. Anticipating this is important for safety. New approach allows more extensive stress-testing ahead of time 👇🏾
Adarsh Subbaswamy
@_asubbaswamy
New preprint w/ (co-first author) @royjamesadams and @suchisaria: "Evaluating Models Robustness Under Dataset Shift" arxiv.org/abs/2010.15100
How can we evaluate *ahead of time* whether or not a model's performance will generalize from training to deployment? 1/ -
Suchi Saria reposted
Alexander D'Amour (@alexdamour.bsky.social)
@alexdamour
Looks like a great way to check whether a model satisfies application-specific robustness criteria.
Gives an efficiently estimable guarantee for worst-case performance across a family of practically relevant settings.
We need more configurable stress testing methods like this!
Adarsh Subbaswamy
@_asubbaswamy
New preprint w/ (co-first author) @royjamesadams and @suchisaria: "Evaluating Models Robustness Under Dataset Shift" arxiv.org/abs/2010.15100
How can we evaluate *ahead of time* whether or not a model's performance will generalize from training to deployment? 1/ -
Suchi Saria
@suchisaria
Hugs everyone! 🤗
Every victory has two sides. I’m very happy about what’s to come. But we need to unite so we can preserve our strength to fight real threats. My takeaways from the last 4 years — we need to become better listeners, drop our judgmental side, find common ground. -
Suchi Saria reposted
ODSC (Open Data Science Conference) AI
@_odsc
It’s time for the #odscwest keynotes, Suchi Saria, PhD (@suchisaria) and Raluca Ada Popa, PhD (@ralucaadapopa)! Listen to their inspirational talks here: attendee.gotowebinar.com/register/5828851298330961166?hss…
-
Suchi Saria
@suchisaria
AI/ML tweeps, here is a must read.
It’s a wonderful article questioning whether we are spending our valuable brain cycles on the right research questions.
If you disagree, I’d love to hear why. If you agree, what can be done to accelerate this? www.technologyreview.com/2020/08/18/1007196/ai-research-m… -
Suchi Saria
@suchisaria
I’d like to sow this seed for AISTATS since reviewing’s abt to begin—I hope this year we will accept every paper that’s a good idea w/o worrying abt acceptance rates.
We are spending VALUABLE research time recycling good papers rejected for idiosyncratic reasons. Let’s fix this.
Suchi Saria
@suchisaria
RT above if you support this so we can get our community to make changes. I felt we could have done better w/ NeurIPS and I’m hoping the OC for @Aistats2020 will set an example. Thanks!
Reviewers, OC can’t do this w/o your help—eg tips on good reviewing: 2020.emnlp.org/blog/2020-05-17-write-good-reviews -
Suchi Saria reposted
TEDMED
@TEDMED
AI Health Pioneer Suchi Saria (@suchisaria) tells us that data is all around us and together with machine learning, we can save more lives. In her powerful #tedmed 2020 Talk, Suchi takes us on a journey of perseverance and ingenuity.
www.tedmed.com/talks/show?id=771271 -
Suchi Saria reposted
Daniel Coelho de Castro
@dccastr0
📢 We're delighted to have @suchisaria as a surprise guest at our #causemic round table, along with our keynote speaker @eliasbareinboim and co-organisers @Imaging_Gen, @kristin_linn, @GlockerBen, and @pwnic!
Tune in at 14:45 UTC 8 Oct #miccai2020
sites.google.com/view/causemic/ -
Suchi Saria reposted
Daniel Coelho de Castro
@dccastr0
Fabulous keynote by @suchisaria at #miccai2020, highlighting the importance of modelling causal relationships for robust and trustworthy #ml4h! For those wondering how to apply this sort of reasoning to medical imaging, we give some pointers in our paper: www.nature.com/articles/s41467-020-17478-w
-
Suchi Saria reposted
MICCAI Society
@MICCAI_Society
Today's keynote at 14 UTC will be given by @suchisaria on building safe and reliable AI - do not miss it! #miccai2020
-
Suchi Saria
@suchisaria
Congrats Princeton!
-
Suchi Saria reposted
Rod Rivera
@rodriveracom
The talk with @suchisaria at @SMILESkoltech is super relevant for academics and industry practitioners. Everyone should be paying attention to robust models and data shifts!
Evidently AI
@EvidentlyAI
What are the similarities between machine learning and reliability engineering? Great talk by @suchisaria on how to develop models that are safe to use and robust to data shifts. Available in recording:
-
Suchi Saria
@suchisaria
Similarly, painful to hear strong feedback on “the right way” to develop/evaluate methods fro clinical reviewers. Often they’re going off common practice or checklists & not using judgement to evaluate the study at hand! Luckily, good editors can save the day!#prematureregulation
Lauren Oakden-Rayner 🏳️⚧️
@DrLaurenOR
This feels like whinging but being asked for references re basic medical concepts is a real pain when I submit to medical ML venues.
Last year was "serious diseases are rarer than benign ones."
This year it'll be "image features of pathology vary more than features of anatomy." -
Suchi Saria
@suchisaria
This is going to be my best presentation yet on the topic! =] As ML becomes part of our every day lives, understanding how to remove failure modes is critical. We've written several papers but it's taken many iterations to make this topic accessible to a broad audience.
Join!
Skoltech
@skoltech_news
⏲️ 20:00 (GMT+3).
Causality and Increasing Model Reliability:
Learning models that are safe and robust to dataset shifts — @suchisaria
👉 www.youtube.com/watch?v=0EP_wR0YT4A -
Suchi Saria reposted
Bayesian Health
@bayesianhealth
Very nice overview article on where do we stand with the development of COVID-19 vaccines: www.washingtonpost.com/graphics/2020/health/covid-vaccine…
-
Suchi Saria
@suchisaria
What’s the death rate of patients on a ventilator w/ COVID-19? A simple question but the answers have been all over the place so we wrote a review.
Most surprising finding: beyond age, comorbidities, degree to which a hospital’s short on resources significantly impacts mortality
Bayesian Health
@bayesianhealth
The mortality rate for intubated #covid19 patients has been reported as high as 97% and as low as 17%.
We performed a review of the literature to find the actual mortality rate, and learn why the reports vary so much.
medium.com/@BayesianHealth/the-hidden-risk-factor-fb2383c…
Let us know what you think! -
Suchi Saria
@suchisaria
Listen to Judea here :-)
Judea Pearl
@yudapearl
These are important papers that should be read by all ML folks who talk about “transfer learning,” “domain adaptation,” or “lifelong learning”. I wonder if the Epi folks are aware of these general methods for "counterfactual predictions" @barbradickerman @_MiguelHernan
-
Suchi Saria
@suchisaria
Clever idea for dramatically reducing the cost of testing and making frequent tests possible:
Ziad Obermeyer
@oziadias
Imagine testing everyone for COVID, every day.
Sure it’d be expensive—but we’d diagnose infections in real time, and keep everyone else safe.
We think we have a way to make this possible… at <$5/person/day.
A thread on our new working paper:
ziadobermeyer.com/wp-content/uploads/2020/07/GroupTesting… -
Suchi Saria
@suchisaria
Telehealth & virtual care are going to be fastest growing segments of healthcare thanks to COVID so it’s an excellent time to learn & be a part of this community!
Huge thanks to the American #telehealth Assoc for inviting me to give the closing keynote. Was a great experience!
American Telemed
@AmericanTelemed
At #ata2020, @suchisaria (a pioneer in #ai) says this about her field: Is it all about hope and promises — or are we stuck in the 1970s? The answer is no; she explains how much progress she's seen in her keynote. Tune i: www.gotelehealth.live/live-stream/15337899/Keynote-Suchi-… #gotelehealth
-
Suchi Saria
@suchisaria
Excellent thread on an extensive but partial list of papers related to algorithmic bias and reliability (COI: includes a couple of my papers). The ideas below are significantly more diverse than the often discussed dataset bias idea.
-
Suchi Saria
@suchisaria
Sad.
People argue fiercely when they’re feeling unheard or threatened. This also leads to banding w/ “their own”, furthering polarization.
America feels so divided even on issues that shouldn’t be divisive. Why? It deeply troubles me. Can we problem solve our way out of this... -
Suchi Saria
@suchisaria
What better way to spend Saturday afternoon than at a workshop on RL in the real world with some really great researchers! (And my 2nd—a panel with rockstar researchers, all female.)
Everything is virtual so you can join at your convenience: sites.google.com/view/RL4RealLife
Yuxi Li
@yuxili99
With great lineup of invited speakers/panelists and moderators: general RL: @edchi @chelseabfinn Jason Gauci, Peter Stone; RL+healthcare: @FinaleDoshi @prasadNiranjani @suchisaria @SusanMurphylab1.
-
Suchi Saria
@suchisaria
Very excited about this. I get invitations in my inbox all the time but this one is special as I'm presenting as the CEO of Bayesian! :=]
Bayesian Health
@bayesianhealth
This Thursday, Bayesian Health's very own CEO, Dr. Suchi Saria, is delivering #ata2020's Keynote! Register at www.xpressreg.net/register/TELE0520/landing.asp?aban=&hke… to learn about the power that #ai has to transform healthcare.
-
Suchi Saria reposted
Daphne Koller
@DaphneKoller
Immigrants founded over half of $1B startups and they or their children founded 45 percent of Fortune 500 companies. These companies create jobs for Americans, not take jobs from Americans. www.forbes.com/sites/stuartanderson/2018/10/25/55-of-amer… www.newsweek.com/immigrant-founded-fortune-500-companies-…
-
Suchi Saria reposted
Rasu Shrestha MD MBA
@RasuShrestha
"It's unprecedented how many times we're using the word unprecedented."
@vmocherla @a16z
Loved today's #covid19 #strategy #investment @zoom_us huddle, as we discussed the "next normal". -
Suchi Saria
@suchisaria
Oh yea, that's fun... and I have the best ideas so that helps my grad students become most happy and productive!
How are you feeling being on the receiving end? ;=] -
Suchi Saria
@suchisaria
Need help re: google scholar:
It links to arxiv version of an article long after it’s been published in a peer-reviewed source. I’ve tried to manually edit link on my scholar page but that doesn’t change listing when you search. How can I fix this? Eg scholar.google.com/scholar?hl=en&as_sdt=0%2C21&q=Treatmen…
Suchi Saria
@suchisaria
The issue is that for some conferences, their formatting isn’t in line with what scholar expects so their papers don’t automatically index correctly. GS can work with conf organizers to fix this. When this isn’t possible, I haven’t found a resolution yet.
-
Suchi Saria
@suchisaria
I’ve heard much debate about whether timing of meals (morning vs evening heavy) is important or is it just total calorie intake plus the length of the eating window. Hard to conduct well-controlled studies but this one presents pretty compelling data: onlinelibrary.wiley.com/doi/full/10.1002/oby.20460 1/
Suchi Saria
@suchisaria
For 93 patients randomized to two arms (breakfast vs dinner heavy) eating similar total calories at similar meal times, the breakfast heavy group lost significantly more weight & waist size. They do account for non-compliance. 2/
Suchi Saria
@suchisaria
One issue I see is that they don’t report % deviations within each group & b/c the D-group reported more hunger, they may be deviating more even if not enough to the point of being asked to drop out. But overall looks compelling. Did I miss anything?
Trying to fix life habits! -
Suchi Saria
@suchisaria
Got to start Sunday@7:30am with my #iclr2020 talk on "safe and reliable ML: critical for building trust."
Slides: www.dropbox.com/s/6lwt8qudtgmch5h/202004--ICLR-Realworld.…
Fun panel w/ @SusanMurphylab1 @nunuska @future_andi & Samantha K discussing real-world apps that also motivated good technical research. -
Suchi Saria reposted
Ciarán Gilligan-Lee
@quantumciaran
Causal inference can have a huge impact in healthcare, drug discovery, & beyond - if used correctly. Check out my latest for
@newscientist , featuring discussions with @eliasbareinboim, @Susan_Athey, @suchisaria, @KiraRadinsky, @yudapearl, @AdrianRaftery1 www.newscientist.com/article/mg24632790-700-correlation-o… -
Suchi Saria reposted
Dr Ellie Murray, ScD
@EpiEllie
This is probably one of the most important tweets you’ll see today. Those headlines saying 88% of people on vents die? They come from a study that excludes anyone who’s *still on a vent* after ~5 days. This is not an appropriate way to calculate survival 🤦🏼♀️
John W Scott, MD MPH
@DrJohnScott
Look at Table 5. First it's amazing to see a paper w/ >1,100 vented #covid19 pts. But of these 1151 vented pts, *at a median follow-up time of 4.5d* only 38 went home, 282 died, while 831 were still hospitalized. 282/(38+282) = 88% mortality. But thats *quite* misleading👇(4/x)
-
Suchi Saria
@suchisaria
Interesting article from UW around the impact of adding an “occasional visit to a friend” over current social distancing.
TLDR; we don’t need super spreaders. An occasional visit takes away a lot of the benefit of social distancing: www.washington.edu/news/2020/04/13/just-one-friend-covid-19/ -
Suchi Saria reposted
Eliza Strickland
@newsbeagle
There's a lot of hand-wavy talk about #ai to combat #coronavirus. Here's a place where it could really be useful: helping hospital docs triage #covid19 patients. My latest: spectrum.ieee.org/the-human-os/artificial-intelligence/me… Interviews with @suchisaria and researchers at @UChicagoMed and @StanfordMed
-
Suchi Saria reposted
Ben Adida
@benadida
1/ Last night, I spent some quality time with the Apple docs on the new contact tracing protocol and APIs they and Google are preparing.
I'm quite optimistic about this effort. Here's why. -
Suchi Saria
@suchisaria
I can’t believe healthcare operates the way it does:
In other industries like finance & retail, within mins or hrs, market moves are discovered & accounted for in future decisions! Big money is on the line so these people are also careful!
In HC, lives are on the line yet... -
Suchi Saria
@suchisaria
Thank you to all the people working so hard to keep New York safe and all the patients cared for! 🤗❤️
Archive: Governor Andrew Cuomo
@NYGovCuomo
We’re going to get through this because we are #newyorktough. @convictsnyc
x.com/convictsnyc/status/1246110693402828800/video/1
Suchi Saria
@suchisaria
1/ We have now 33 Hubei (states with higher prevalence than Hubei at the time of its shutdown) yet most haven’t implemented strict social distancing measures yet: medium.com/@tomaspueyo/coronavirus-out-of-many-one-36b886…
Very thoughtful read, highly recommend it. -
Suchi Saria
@suchisaria
Have been on the phone with friends and family all over India. They went from not caring and celebrating Holi (the festival of colors) to a lock down in 1 week! The worrisome things I learnt:
Post not available here, view on X
Suchi Saria
@suchisaria
5/ Mumbai is as densely packed as NYC & India started lock down only a wk ago w/ no infra for testing or clear directives for transport. India may already be in worse shape & we won’t get real death count estimates near term. Need more prescriptive policies+education there.
Suchi Saria
@suchisaria
6/ There are HUNDREDS of THOUSANDS of people on our planet who overnight lost their jobs, are food-deprived, & their health is in danger. This is very real & very sad.
If nothing else, let’s make sure we don’t spend time complaining & use it for good.
www.bloomberg.com/news/articles/2020-03-30/-we-will-starv… -
Suchi Saria reposted
Nicholas A. Christakis
@NAChristakis
Let’s talk about what happens if you get COVID19 and recover. Are you immune to the disease? How long does the immunity last? And what does that mean for your life and for the public health and economy of our society? 1/
-
Suchi Saria
@suchisaria
The paper we have been waiting for : an analysis of how temperature and humidity influence COVID-19 transmission rates.
poseidon01.ssrn.com/delivery.php?ID=635090098026018005095…
Suchi Saria
@suchisaria
The paper selects data from periods prior to the quarantine to control for confounding due to how this intervention might influence spread differently across cities. Higher temperature & humidity does seem to lower the spread factor when seeing data across cities in China
Suchi Saria
@suchisaria
This analysis looks at spread across different countries. I trust this graph less because test availability has also been worse in countries below the equator. Wonder if the outliers (High R in high temp/humidity regions) are explains by age or any other factor?
-
Suchi Saria reposted
John Arnold
@johnarnold
Short term social distancing won’t completely kill CV & long term distancing not practical. This thread is best explanation I’ve seen about long game of CV. Short answer is that most people will eventually get it but how and when drives outcomes. Written from U.K. perspective.
Prof Ian Donald @professoriandonald.bsky.social
@iandonald_psych
1. The govt strategy on #coronavirus is more refined than those used in other countries and potentially very effective. But it is also riskier and based on a number of assumptions. They need to be correct, and the measures they introduce need to work when they are supposed to.
-
Suchi Saria
@suchisaria
If you need a good laugh, this is it!
I wonder how badly the hosts of MSNBC are regretting this! -
Suchi Saria reposted
ACM Education & Learning Center
@acmeducation
Mar 25 #acmtechtalk "The Decision-Making Side of Machine Learning: Computational, Inferential, and Economic Perspective," w/Michael I. Jordan. event.on24.com/eventRegistration/unavailable?&eventid=221…
@JeffDean @smolix @etzioni @erichorvitz @lexfridman @DaphneKoller @pabbeel @ShamKakade6 @suchisaria @aleks_madry -
Suchi Saria reposted
-
Suchi Saria reposted
aydao
@AydaoAI
The AI makes them blink and gives them teeth... 😬 (take 2)
-
Suchi Saria
@suchisaria
This is such a beautiful video.
We are so divided and polarized in the US right now. We can learn from this: youtu.be/jD8tjhVO1Tc -
Suchi Saria
@suchisaria
My favorite version of #ted — some of the best science and discovery talks. Very nervous about delivering my 9 pager but overall, can’t wait!
It’s in Boston this year and being live streamed. Join!
www.eventbrite.com/e/tedmed-live-stream-tickets-96772929437 -
Suchi Saria reposted
François Chollet
@fchollet
A deep learning model that can detect COVID-19 pneumonia from CT scans. When used by radiologists, the model can cut reading time by 65%. This can help take pressure off frontline radiologists if/when the number of cases explode.
www.medrxiv.org/content/10.1101/2020.02.25.20021568v1
(This work uses Keras) -
Suchi Saria
@suchisaria
So honored to be invited as an associate editor and to serve alongside many of these brilliant researchers in advancing critical care! @SCCM
Sapna Kudchadkar, MD, PhD
@SapnaKmd
The @CritCareExplore Editorial Board - the future of all @SCCM journals. Congratulations!
-
Suchi Saria reposted
Critical Care Medicine
@CritCareMed
Too Many Definitions of Sepsis: Can #machinelearning leverage the #ehr to increase accuracy & bring consensus?
@SCCM #critcaremed @suchisaria @KatharineEHenry #sepsis
Link: journals.lww.com/ccmjournal/Fulltext/2020/02000/Too_Many_…
Fig: Applicability of Automated Sepsis Identification Methods -
Suchi Saria
@suchisaria
Humans pranking their cats — this video is just too good!
Watch the last one in particular...hahaha -
Suchi Saria reposted
Hemant Pathak
@hpinreston
Great to partner with @suchisaria from Johns Hopkins University and @jessicazeaske to discuss Responsible AI in Healthcare at #hdpalooza this morning. #msftinhealth
-
Suchi Saria
@suchisaria
I’d add-> healthcare is not a zero sum game. Wouldn’t it be great if we could provide care twice as fast at half the cost and made it possible for everyone to live twice as long? We need all the help we can get, from both human experts and machines!
-
Suchi Saria
@suchisaria
Friends, I’m giving a talk at this DC/policy focused meeting: www.academyhealth.org/page/2020-hdp-agenda
I’m looking for egs of tasteless #ai/ML algorithmic research or poor implementations that could have harmful consequences, to highlight what they did wrong & what right might have looked like.
Suchi Saria
@suchisaria
Here is an example:
🇵🇸 Kai Daniel Moore
@bruise_almighty
I had to get a background check for my job, and it turns out the report is a 300+ page pdf of every single tweet I’ve ever liked with the work “fuck” in it.
Enjoy your dystopian bs! *waves* -
Suchi Saria
@suchisaria
My obs: Many DL/CI/AI debates sparked by non-expert twitter provocateurs & core researchers feel the pressure to respond!
Our productive scientists ought to be helping make research progress! Cur back & forth presents a hostile image often rehashing the same topic. Am I wrong?
Suchi Saria
@suchisaria
CI is looking to draw causal conclusions (many techniques can he applied to estimate said effects inc neural networks)
DL uses variants of a particular class of models to fit functions (used for many goals inc causal inference).
They’re not at odds or 1 a superset of the other. -
Post not available here, view on XPost not available here, view on X
Suchi Saria
@suchisaria
CI is looking to draw causal conclusions (many techniques can he applied to estimate said effects inc neural networks)
DL uses variants of a particular class of models to fit functions (used for many goals inc causal inference).
They’re not at odds or 1 a superset of the other. -
Suchi Saria
@suchisaria
This study’s caused a stir. My take:
-regression to the mean is a common flaw in pre/post studies; this intervention was esp vulnerable b/c it selected patients at the peak. This could have been ~corrected for(even w/o RCT)
-but has standard of care also improved sig over time? -
Suchi Saria
@suchisaria
Just noticed our cookbook shelf has one VERY odd resident!
(Organized by our movers a year ago.) -
Suchi Saria
@suchisaria
This is no different than how other non-AI diagnostic/screening tests have been introduced before:
analyze the safety and efficacy — ie, downstream consequences of over treatment and missed diagnosis before introducing a protocol that’s deployed at scale
Nikhil Krishnan
@nikillinit
Some big AI papers coming out (Google Mammogram, Moorfields + alzheimers)
1) Should we be doing screening at scale if we don't understand the underlying disease and if we can do anything with what we find
2) Should retrospective studies be allowed to go into clinics -
Suchi Saria
@suchisaria
This isn’t surprising.
DL uses a class of models; I think of the rest as ML—learning frm small sample, correcting 4 bias, inferring causal effects...
But at the height of DL fever, DL got branded as a separate field & known learning principles got reinvented w/ DNN in the title
Yann LeCun
@ylecun
Some folks still seem confused about what deep learning is. Here is a definition:
DL is constructing networks of parameterized functional modules & training them from examples using gradient-based optimization.... www.facebook.com/722677142/posts/10156463919392143/ -
Suchi Saria
@suchisaria
If you love puzzles and decoding magic, and want to be entertained, visit the @nytimes crossword constructor @davidkwong ‘s magic show.
It is absolutely brilliant! Mind blown. (And, the perfect idea for a NYC date!) -
Suchi Saria
@suchisaria
Imagine discussing machine learning and medicine in Bermuda with other top researchers across informatics, policy, clinical translation, and ML. That's what this meeting is about!
Abstracts due soon. If you have questions, send me a note.
NEJM
@NEJM
NEJM Group is supporting SAIL 2020. Call for abstracts now open. Submit by Dec 20. #machinelearning #clinicalinformatics #computationalmedicine sail.health/event/sail-2027/participate/
-
Suchi Saria reposted
Kate Crawford
@katecrawford
Finally, a paper on why "AI for good" is an empty phrase without a theory of change. What is "good" is never articulated in the rush to tech solutions, while alternative reforms are overlooked. Read @benzevgreen's piece before it blows up at @NeurIPSConf www.benzevgreen.com/wp-content/uploads/2019/11/19-ai4sg.pdf
-
Suchi Saria reposted
Sri 'Maverick' Srihari
@srigsri23
Watch “Panel: How Connected Data Can Power a Learning Healthcare System” on #vimeo vimeo.com/372294350
Valuable point by @suchisaria: care providers should see the system improving their productivity, helping them do something better esp when already have lots of things to do -
Suchi Saria
@suchisaria
Nice paper by @ravi_b_parikh et al. w/ examples disentangling social vs statistical bias & ideas for how (well-designed) AI/ML tools can help mitigate social bias: jamanetwork.com/journals/jama/fullarticle/2756196?guestac…
We can't solve a problem we don't understand. To address bias, we need a cleaner taxonomy.
Suchi Saria
@suchisaria
Related: Published in Biostatistics this week, this article overviews learning methods that correct for statistical biases commonly observed when developing tools from real-world EHR data.
academic.oup.com/biostatistics/advance-article/doi/10.109… -
Suchi Saria reposted
Harlan Krumholz
@hmkyale
Why would we use #ml to predict risk of kidney injury based on contrast exposure during stent placement? Because the relationship is complex. Non-linear and nuanced. We can individualize the predictions. jamanetwork.com/journals/jamanetworkopen/fullarticle/2755869 @JAMANetworkOpen @jbmortazavi
-
Suchi Saria reposted
Adarsh Subbaswamy
@_asubbaswamy
Highly recommend checking out this issue.
@suchisaria and I discuss problems w/ differing training and deployment conditions, and how causality gives us a language/tools to express and identify the problem, and tools to develop new robust algorithms.
academic.oup.com/biostatistics/advance-article/doi/10.109… -
Suchi Saria reposted
Rahul Rajkumar
@RahulRajkumar11
Today is a tipping point for value-based care in #nc. @AtriumHealth and @NovantHealth are now part of #bluepremier, our value-based program that holds health systems and @BlueCrossNC jointly accountable for better health outcomes and lower costs. mediacenter.bcbsnc.com/news/blue-cross-nc-atrium-health-a…
-
Suchi Saria reposted
Just another south FL dude
@drsxr
Excellent HBR article citing brilliant @suchisaria who I’d cite too (& who’s work I only partly understand) - #ai #radiology should consider some of this. Cc: @janbeger @FelipeKitamura @alexandrecadrin @judywawira @DrGMcGinty @quantrad @RasuShrestha
Mash Zahid
@MashZahid
"Using AI to Understand What Causes Diseases" (HBR), highlights importance of causal AI and counterfactual models. Includes context with useful links to Px adherence, re-admission & disease predictions, which were bleeding edge projects mere few years ago. hbr.org/2019/11/using-ai-to-understand-what-causes-diseases
-
Suchi Saria reposted
Farzad Mostashari
@Farzad_MD
Here's the thing.
1/ The perception of Google culture is that no-one curbs the curiosity of engineers
Google can sign a BAA, but they have to convince people that they actually have controls in place to ensure that the data is only being used for the purposes of the agreement
Christina Farr
@chrissyfarr
Google: “It’s understandable that people want to ask questions about our work with Ascension.”
cloud.google.com/blog/topics/inside-google-cloud/our-part… -
Suchi Saria
@suchisaria
Aww, thx!
My favorite meetings are when the discussants & speakers prior to me motivate new talk threads.
Vincent Liu from @aboutKP led to my creating of this slide asking when do we need high quality models vs when are simple rule-based/epic-style predictive systems enough? -
Suchi Saria
@suchisaria
All of the above.
Bill Padula, PhD
@DrWmPadula
How will #machinelearning #artificialintelligence best support #research in #healtheconomics and #healthcare?
@DrBlytheAdamson @DrPJonsson @brootmcq @joeymattingly @AnkurFactorial @Basucally @HEvigilante @ISPORorg @drnigam @suchisaria @maartenijzerman @djvanness @atulbutte -
Suchi Saria reposted
Jonathan H Chen MD PhD
@jonc101x
Privileged to share the stage with @sallyokun Mildred Cho @embimd @miakeeys and my co-authors @AndrewLBeam @suchisaria Eneida Mendonca who have taught and inspired me in our panel and chapter on the potential #unintendedconsequences of #ai in #medicine
-
Suchi Saria
@suchisaria
The devil is in the details though? — how will they enable caring for health and wellness and over time how does that translate into savings and outcomes. @onemedical is just a doctor’s office but it’s led to a big change in expectations for how individuals access everyday care.
Matthew Holt
@boltyboy
Apparently Haven (the @AtulGawande1 @amazon - @chase- @BerkshireHathEC health employer thing) is out with its first offering and it's a....
...1990s style HMO
www.geekwire.com/2019/amazon-jpmorgan-roll-new-health-ins…
Suchi Saria
@suchisaria
My observation about healthcare innovation — we are obsessed with new concepts but sometimes are too impatient to execute well on hard ideas. Attention to detail is what generates beautiful execution and it requires bucket loads of patience (eg a lot of great consumer tech).
-
Suchi Saria reposted
Sarah Candler
@sarahgcandler
“Shall we play a game?”
#artificialintelligence in Medicine: the future of healthcare
- AI as a second set of eyes
- discovering symptoms/trends docs can’t see
- assist with scaling diagnostic care (humans are hard to scale)
@Texas_ACP with @suchisaria -
Suchi Saria reposted
National Academy of Medicine
@theNAMedicine
Join us for a panel on using AI for care delivery with with @suchisaria (@JohnsHopkins), Amy Abernethy (@HHSGov), @atulbutte (@UCSF_BCHSI), @amyleecp3 (@psjh) & @m_sendhil (@ChicagoBooth), starting now! Webcast: nam.edu/event/the-evolution-of-technology-in-biomedical-s…
-
Suchi Saria
@suchisaria
Beautiful article by @Carolynyjohnson in the @washingtonpost on bias in algorthmic systems done by @oziadias and team.
As I describe within, a key opportunity I see is the ability to now measure and correct for these biases because the digital infrastructure exists!
carolyn johnson
@Carolynyjohnson
The machines are here to decide if you get that job interview www.washingtonpost.com/technology/2019/10/22/ai-hiring-fa… @drewharwell or that follow-up medical appointment www.washingtonpost.com/health/2019/10/24/racial-bias-medi…
-
Suchi Saria
@suchisaria
Friends: I'm speaking this Monday at the National Academy of Medicine annual meeting on AI/ML. I need your help.
What are the most impactful results you've seen for #ai / ML in health? Please add rationale for selecting.
Doesn't have to be implemented at scale yet. Tx! -
Suchi Saria
@suchisaria
This is only one of many ways to screw up.
My experience: the more deeply you understand medicine & ML —> the longer it takes to write a paper.
Effort required for careful dx of whether a software is doing what’s intended is 10x the effort to write a paper w/ AUC reporting!
Lauren Oakden-Rayner 🏳️⚧️
@DrLaurenOR
"Improving Medical AI Safety by Addressing Hidden Stratification" is new research by @jdunnmon (@Stanford) and myself (@TheAIML). We argue HS can explain failures of pre-clinical #medical #ai testing, and propose an achievable regulatory solution.
Blog: lukeoakdenrayner.wordpress.com/2019/10/14/improving-medic…
Suchi Saria
@suchisaria
Other examples if you're curious: proceedings.mlr.press/v89/subbaswamy19a/subbaswamy19a.pdf and proceedings.mlr.press/v89/schulam19a/schulam19a.pdf
-
Suchi Saria
@suchisaria
Impressive how much back & forth the deconfounder paper has been a part of. In this case I agree with Ogburn et al.
(There is a cool idea there about using predictive checks to help rule out/in assumptions to determine identifiability but overall original claims were too broad.)
Edward Kennedy
@edwardhkennedy
Woah. What a first sentence for an abstract:
"The premise of the deconfounder method proposed in "Blessings of Multiple Causes"... is incorrect."
arxiv.org/pdf/1910.05438.pdf -
Suchi Saria
@suchisaria
Such an interesting set of thoughts. And easily as productive as investing in AGI. I’d pay if these tools existed.
@michael_nielsen @andy_matuschak
Andy Matuschak
@andy_matuschak
Tools for thought are a beautiful idea—inventions which can “change the thought patterns of an entire civilization.” But that’s a 30 year old quote. Why are they so hard to make? @michael_nielsen and I try to answer that question and suggest paths forward: numinous.productions/ttft
-
Suchi Saria
@suchisaria
Nice example of how AI will up-level the workforce by increasing efficiency and effectiveness
www.statnews.com/2019/10/01/an-ai-startup-led-by-23andme-… -
Suchi Saria
@suchisaria
Had lunch with a healthcare investor—awesome guy but their portco had sold him on all sorts of stuff ML/AI can do that it can’t quite do yet & its subtle if you’re a non-expert to evaluate where issues arise. He is helping them scale; I worry, screwups in HC can hurt lives.
-
Suchi Saria
@suchisaria
Nice Leora! Key takeaways:
- Get your IRB squared away so that the approvals don’t hinder progress
- Access to IT to add *and remove* interventions
- Carefully measure to understand whether something is working (independent of randomization)
Bring a *scientific mindset* to QI.
Leora Horwitz
@leorahorwitzmd
Excited to see work I and many colleagues @nyulangone have done to create a learning health system through randomized quality improvement work now in print at @NEJM. Special thanks to @jones_prof @devin_mann @PaulTestaMD & many non twitter peeps. www.nejm.org/doi/full/10.1056/NEJMsb1900856
-
Suchi Saria
@suchisaria
I love this. What do you think?
Atul Butte (In Memoriam)
@atulbutte
Participating in my final @NICHD_NIH Council meeting, after serving for 4 years. My final thoughts are here.
-
Suchi Saria reposted
Join Y. Luh, MD (he/him/dude/他)
@j_luh
"How does the #physician know when to trust the machine or not?"
@SuchiSaria PhD discusses challenges in using #ai in #healthcare @ #astro19 Keynote address #radonc #digitalhealth -
Suchi Saria
@suchisaria
Great speakers. Join if you’re in NYC!
@haldaume3 who’s organizing this? Learning policies from retrospective datasets is all about ML & causal inference. @eliasbareinboim & @_asubbaswamy both doing amazing work are nearby—off-policy RL folks will benefit from their perspective.
Hal Daumé III
@haldaume3
In NYC area and interested in reinforcement learning?
Come to RL Day at @MSFTResearch on Oct 3!
Free to register!
Awesome speaker lineup: Joelle Pineau, @smdvln, Christopher Amato, Mengdi Wang, Philip Thomas, @real_asli, Geoff Gordon and @FinaleDoshi!
www.microsoft.com/en-us/research/event/reinforcement-lear… -
Suchi Saria
@suchisaria
Behind a paywall but this sounds exciting. Looking forward to reading it. If anyone ends up reading it sooner, curious to hear about how they designed the reward functions, the input dimensionality of the problem, and the degree to which modeling nonlinearity was critical.
Eric Topol
@EricTopol
There's been a lot of buzz about #ai for drug discovery. But I think this is the best case yet: All in 46 days!
www.nature.com/articles/s41587-019-0224-x
@InSilicoMeds @uoft @VectorInst @biogerontology @A_Aspuru_Guzik and colleagues
@NatureBiotech #deeplearning -
Suchi Saria
@suchisaria
Wow—my genes are working in my favor? “Fu says researchers have found that short sleepers tend to be more optimistic, more energetic and better multitaskers.
They also have a higher pain threshold, don’t suffer from jet lag and some researchers believe they may even live longer”
Meredith Turner
@UC_Newsroom
After a decade of searching, the @ucsf scientists who identified the only human gene known to promote “natural short sleep”—lifelong, nightly sleep that lasts just 4-6 hrs yet leaves individuals feeling fully rested—have discovered a second @neurocellpress www.ucsf.edu/news/2019/08/415261/after-10-year-search-sci…
-
Suchi Saria
@suchisaria
Love it!
Policy recommendation to increase use of real-world patient outcome data with #ml #ai to bring more rational outcome-based pricing to drugs and reduce cost of drug trials.
Bipartisan Policy Center
@BPC_Bipartisan
A special thanks to former Senate Majority Leader Dr. @bfrist, as well as former @US_FDA commissioners, Drs. @califf001, Andrew von Eschenbach, and Mark McClellan, for all of the guidance in writing our 'Expanding the Use of Real-World Evidence' report! bpcdc.org/2zaJzJn
-
Suchi Saria reposted
Eric Topol
@EricTopol
An important roadmap for #ai in medicine w/ a call "Do No Harm" @NatureMedicine by @jennawiens @suchisaria and many leading lights in the field, including @UofTCompSci @uoftmedicine @VectorInst www.nature.com/articles/s41591-019-0548-6
-
Suchi Saria
@suchisaria
Entrepreneurs taking matters into their own hands to fix the American health system — nice!
Ciitizen designs score cards for rating health systems on how good they are about giving patients back their data when asked! #gociitizen
cc @a16z @vijaypande @ciitizencorp
Matthew Holt
@boltyboy
Guess who gets 1 star (the worst you can get) @MayoClinic @StanfordMed @sloan_kettering @UCSFHospitals
-
Suchi Saria
@suchisaria
The benefit of being a senior researcher is that you have so much riches to go around, you can be generous. I’ve not been around for all that long but I’ve actively felt this way for a few years now.
Wouldn’t it be great if we had a service that reminded folks in a polite way? -
Suchi Saria
@suchisaria
Love that the NHS is doing what’s right—using their data to level up care—& I hope systems in the US will follow.
Most top places for ad placement or stock trading can’t imagine using only a manual process, yet we do this for the most critical decisions in our life — our health!
Transforming health and care
@NHSTransform
Today we can announce that we're going to set up an NHS AI Lab - in partnership with our pals at the Accelerated Access Collaborative.
We're going to focus on practical AI tools that will make a difference to patients, and we're going to embed data ethics in all of our work. -
Suchi Saria
@suchisaria
Twitter peeps, am a prof at Hopkins in engineering and health policy. I’m in NYC on sabbatical and looking to meet health policy / health data (nerds) living here. My thought was to host a wine and cheese as a get to know you. Would love intros/pointers to your favorite folks.
Suchi Saria
@suchisaria
Loving all the recommendations! Looking fwd! If you haven't yet, please DM me your email so I can send along details (won't be divulged elsewhere so don't have to worry about spam). I'm thinking a Saturday afternoon. Given the size, will likely move it to a garden patio at a bar.
-
Suchi Saria
@suchisaria
Slides from my keynote at #uai on "Safety Challenges with Black-Box Predictors & Novel Learning Approaches for Failure Proofing"
It contains many ideas at the intersection of causal inference & #ml with a focus on ensuring safety / reliability
www.dropbox.com/s/jxefbgjnyu88luv/2019-UAI-FailureProofin… -
Suchi Saria reposted
Judea Pearl
@yudapearl
The idea of "surgery estimators" is ingenious, it would not occur to me that you can get extra mileage on top of "pruned estimators". However, is Figure 2(b) the simplest example to show this extra mileage?? Would c-equivalence help here? ucla.in/2NbJUVq #bookofwhy
-
Suchi Saria reposted
Judea Pearl
@yudapearl
1/ I just read arxiv.org/pdf/1905.11374.pdf and I agree with @suchisaria. Anyone concerned with stability (or invariance) should start with this paper to get a definition of we are looking for and why. Arjovsky etal paper should be read with this perspective in mind, as an attempt to
Ferenc Huszár
@fhuszar
I wrote a post on IRM using more of the language of conditional independence that you might find interesting:
www.inference.vc/invariant-risk-minimization/ -
Suchi Saria
@suchisaria
Tom, I hope this is the case but I still see reluctance in the community towards stating what’s true in the world (this goes against ML instinct of de novo) & desire to clearly understand whether what’s inferred is correct (because beating SOTA is a common yardstick).
Thomas G. Dietterich
@tdietterich
Machine Learning is at the cusp of a "causality revolution" driven by intellectual debt and the framework of @yudapearl. An important recent paper: Invariant Risk Minimization arxiv.org/abs/1907.02893
Two more pioneers in this effort: @eliasbareinboim and @suchisaria -
Suchi Saria reposted
Judea Pearl
@yudapearl
Great! I was't aware of these two papers. Now we know that searches for "stability" and "invariants" are on firm scientific grounds: When we suspect a certain relation is "stable" we can check and see that it is truly stable and what conditions will make it unstable.#bookofwhy
-
Suchi Saria reposted
Ohad Levinkron
@OhadLevinkron
Science is hard!
Sunset at Tel Aviv with @suchisaria and @ShimonShei -
Suchi Saria reposted
Adarsh Subbaswamy
@_asubbaswamy
We're developing a (graphical) language for expressing assumptions about stable correlations (invariant from train to deployment) and a hierarchy of prediction relations (can even be counterfactual!) that will generalize. (proceedings.mlr.press/v89/subbaswamy19a/subbaswamy19a.pdf & arxiv.org/pdf/1905.11374.pdf). 1/2
Thomas G. Dietterich
@tdietterich
A spurious correlation is one that is present in the training (and validation data) but absent in deployment. Standard ML algorithms find these. Yes, causal model assumptions, if correct, can prevent this.
-
Suchi Saria reposted
Thomas G. Dietterich
@tdietterich
I am a big fan of Adarsh's work and recommend these articles.
Adarsh Subbaswamy
@_asubbaswamy
We're developing a (graphical) language for expressing assumptions about stable correlations (invariant from train to deployment) and a hierarchy of prediction relations (can even be counterfactual!) that will generalize. (proceedings.mlr.press/v89/subbaswamy19a/subbaswamy19a.pdf & arxiv.org/pdf/1905.11374.pdf). 1/2
Thomas G. Dietterich
@tdietterich
A spurious correlation is one that is present in the training (and validation data) but absent in deployment. Standard ML algorithms find these. Yes, causal model assumptions, if correct, can prevent this.
-
Suchi Saria
@suchisaria
HAHA
The power goes out because of a tree crash at my friend’s house.
Her little daughter looking anxious goes: mommy, how are we going to know what the weather is because Alexa is not speaking?
#firstworldproblems -
Suchi Saria
@suchisaria
100%
The intersection of bio/health+data is where the most exciting action is/will be over the next few decades.
Meg Tirrell
@megtirrell
.@WalterIsaacson: The first half of the 20th century was shaped by physics. The second half by information technology & microchips. "My belief now is that the technology that will affect the first half of the 21st century most is biotech." www.statnews.com/2019/07/02/walter-isaacson-crispr-books/
-
Suchi Saria
@suchisaria
I almost never took notes as a student or a prof.
#startup life’s taught me the value of detailed meeting notes, follow ups, expressly stating deadlines, bullet lists, listening carefully, and countless other life skills to be more productive!
Better late than never! -
Suchi Saria reposted
Suhail
@Suhail
Good teammates:
(1) challenge your over optimism
(2) don’t bicker about work style differences
(3) prioritize getting work done
(4) care about the goal over themselves
(5) don’t want to waste time
(6) kill their ego
(7) debate to make a great product not to personally win -
Suchi Saria
@suchisaria
True. In addition, prospective studies require freezing the system & workflow. We have been at it for 2.5+ years & there are so many ways to improve performance & integration that it’s hard to overlook these opportunities.
Deeply understanding machine/team interaction is key.
Julian Hong
@julian_hong
Domain knowledge so important for both accuracy and implementation of #machinelearning models in patient care (or any context). I think a big reason for why we’ve seen so many algorithms but few prospective studies published.
-
Suchi Saria
@suchisaria
Excellent choice of partner in @cuttingforstone Looking forward to these.
And what a shift from 2011 where few in medicine discussed AI/ML to only 8 years later and we have senior physicians evangelizing the field and recognizing the change it will bring.
Eric Topol
@EricTopol
We (@cuttingforstone and I) have a new podcast "Medicine and the Machine" in which we (and our guests) explore the potential of #ai to improve medicine and the patient-doctor relationship
www.medscape.com/viewarticle/914761
the 1st episode is on Presence and is now live @Medscape -
Suchi Saria
@suchisaria
Congrats to the 2019 TR35 innovators! Special congrats to @hima_lakkaraju @red_abebe and Archana Venkataraman who are advancing AI / data science in spades for social good! www.technologyreview.com/lists/innovators-under-35/2019/
-
Suchi Saria reposted
Stanford University
@Stanford
"There are too many people who want credit without responsibility. Too many who show up for the ribbon cutting without building anything worth a damn. Be different. Leave something worthy." -@tim_cook #stanford19
This media couldn't be displayed. See it on the original post. -
Suchi Saria reposted
Atul Butte (In Memoriam)
@atulbutte
.@SuchiSaria: What's really changed is how we fundamentally teach computers… #ucaibio
-
Suchi Saria reposted
Atul Butte (In Memoriam)
@atulbutte
.@SuchiSaria: The same thoughts we have today were highlighted in the @NEJM back in 1970! And we think we are so original! www.nejm.org/doi/full/10.1056/NEJM197012032832305 #ucaibio
-
Suchi Saria reposted
Atul Butte (In Memoriam)
@atulbutte
.@SuchiSaria: "To get things to REALLY work, it takes a LOT of work." #ucaibio
-
Suchi Saria reposted
Atul Butte (In Memoriam)
@atulbutte
.@SuchiSaria: #ai-driven nuisance alerts are going to be the future source of physician burnout, if we aren't careful #ucaibio
-
Suchi Saria
@suchisaria
Stuart Schreiber, co-founder of the Broad uses DNA analysis and realizes his dad wasn’t his real dad... This is such a remarkable story of truth seeking and personal courage by a wonderful scientist—wow!
Thank you for sharing this deeply personal story @SchreiberStuart
Stuart Schreiber
@SchreiberStuart
Without love and encouragement from my wife Mimi, this would have remained a personal diary for the two of us. The response from so many of you has warmed my heart. I am profoundly grateful. www.harvardmagazine.com/2019/07/dna-testing-schreiber
-
Suchi Saria
@suchisaria
Very fun discussions but I am feeling a bit stood up.
They are all formally dressed, especially Sham in a nice red tie w/ matching red socks and a black suit while I am in my jeans & vest. I didn’t know ACM/Turing meeting automatically meant upgrade your clothes!
Association for Computing Machinery
@TheOfficialACM
Tune into the panel "Deep Learning, Reinforcement Learning, and Role of Methods in Data Science" featuring Shirley Ho, Sham Kakade (@ShamKakade6), Suchi Saria (@suchisaria) and Manuela Veloso at the ACM-IMS Data Science Summit: www.acm.org/data-science-summit/livestream
@InstMathStat
Suchi Saria
@suchisaria
I decided to one up @ShamKakade6 but may not have fully succeeded. Here are some photos from the evening! Was a real treat to join the Turing award celebrations. My favorite bit was Geoff’s closing words: don’t give up on an idea you believe in even if people call it silly.
-
Suchi Saria reposted
Yisong Yue
@yisongyue
Super excited about our Real-World Sequential Decision Making Workshop at #icml2019!
Great invited speakers: @EmmaBrunskill , Miro Dudik, @suchisaria & Dawn Woodard!
Many great posters as well!
2pm-6pm on Friday June 14th in Seaside Ballroom!
realworld-sdm.github.io -
Suchi Saria
@suchisaria
Just looking at my calendar and realizing our lab has 9 talks/keynotes/panels in the coming week! Happy Monday!
On the upside, if you’re at @icmlconf, @JNJCares, @UCSFPrecision @uchealth or @TheOfficialACM IMS meeting, come say hello! -
Suchi Saria
@suchisaria
This. It’s so useful and the positive benefits are many.
Jen Heemstra
@jenheemstra
Academia has made me really good at coping productively with rejection and criticism. Once I got past my ego, I realized how much fun it is to learn, grow, and improve. The best part is that this has spilled over into my personal life to make my relationships better too.
-
Suchi Saria reposted
Balaji Lakshminarayanan
@balajiln
Looking forward to our @icmlconf workshop on "Uncertainty & Robustness in Deep Learning" at #icml2019 next Friday! We have 53 accepted papers, and invited talks by @wellingmax, Kilian Weinberger, @suchisaria, @dawnsongtweets, @Dr_Innovation :) More info: sites.google.com/view/udlworkshop2019/home
-
Suchi Saria
@suchisaria
Thought provoking read by @JorgeCondeBio a16z.com/2019/05/30/cure-preexisting-condition-gene-therapy/
When measuring ROI as price to the lifetime value created, therapeutics often seem much more expensively priced compared to preventative diagnostics. Why? This continues to puzzle me.
Jorge Conde
@JorgeCondeBio
1/ Recently $NVS announced a $2.1m gene therapy, the most expensive medicine in history. Gene therapies are potentially curative. How should we price them? How will we pay for them? Could a cure be considered a pre-existing condition? 🧬 💊
a16z.com/2019/05/30/cure-preexisting-condition-gene-therapy/ -
Suchi Saria
@suchisaria
Just learnt that one of my favorite humans @withyouDrWu won the AHRQ excellence in mentoring award: www.ahrq.gov/funding/training-grants/eisenberg.html
Mentoring is hard, feels much harder than being a researcher. And it’s been such the joy & privilege learning from him & growing with him. Congrats Dr. Wu! -
Suchi Saria
@suchisaria
These look like a fun way to spend at least some of this weekend!
-
Suchi Saria reposted
Scott Kupor
@skupor
The challenge of developing drugs for treatment of early-stage cancers www.wsj.com/articles/drug-giant-tries-new-tactic-to-fight…
-
Suchi Saria
@suchisaria
Since 1980 patient complexity has grown & more complex therapeutics/procedures have come to market.
To what extent is it this vs bloat? Healthcare could definitely use better tech/automation to increase efficiency but often poor incentives come in the way (eg fax 4 prior auth!).
Eric Topol
@EricTopol
I graduated medical school in 1979 and have watched the changes in the last 40 years, summarized briefly in this Table from #deepmedicine as medicine/#healthcare became a big business
-
Suchi Saria reposted
Rahul Kashyap, MD, MBA 🇺🇸🇮🇳
@Dr_SRKashyap
#mayoclinicai Symposium- meeting the giants in the field of medical #ai @MayoClinic with @suchisaria @Khanna_S @sumitbhagramd
-
Suchi Saria reposted
Aditya Khurana
@adikhurana
Great opening keynote by Dr. @suchisaria at the #mayoclinicai symposium. Much of medicine doesn’t have a gold standard, so we must learn to triangulate from weak signals to inform care
-
Suchi Saria reposted
Farbod Saraf
@farbodsaraf
The pendulum motion x.com/ZonePhysics/status/1127725386379550721/video/1
-
Suchi Saria reposted
Jeff Dean
@JeffDean
I'm looking forward to speaking about the potential of machine learning for healthcare at this @UofCalifornia-wide event on June 17-18 (My talk is on the 17th)!
Other speakers include @suchisaria and Wendy Nilsen of the NSF, among others. -
Suchi Saria reposted
Jeff Clune
@jeffclune
In the ICLR Debate I introduced an alternate path to general AI: AI-generating algorithms (AI-GAs), and the Three Pillars required to make AI-GAs. Full debate (on how much we should learn vs. build in): youtu.be/APcFQoLKVLA w Tenenbaum, Precup, Kaelbling, @suchisaria Thoughts?
This media couldn't be displayed. See it on the original post. -
Suchi Saria reposted
Shakir Mohamed
@shakir_za
At #iclr2019 thanks to all our panelist @suchisaria @jeffclune doina precup, Josh tenenbaum and Leslie kaelbling for a broad and thought provoking discussion on the role of priors and learning. Question for everyone: what do you think can't be learnt?
-
Suchi Saria reposted
Been Kim
@_beenkim
"Evolutionary approach is great, but don't we not want to hurt people on the way?" - @suchisaria #iclrdebate
-
Suchi Saria
@suchisaria
Thanks @tdietterich!
Integration of ML & causal inference are key to AI's real-world use: (1) for risk analysis & failure proofing to enhance reliability and robustness, (2) for drawing reliable conclusions from observation data. 1/
Thomas G. Dietterich
@tdietterich
Causal inference *is* a part of AI and machine learning. And not surprisingly, some of the best research in causal inference & ML is being done by researchers in medical AI such as @suchisaria.
Suchi Saria
@suchisaria
Finding many hesitate to work on these topics b/c it requires familiarity w/ many areas. Adarsh & I put together a ACM tutorial that discusses robustness/reliability and how causal inference & ML play a role: arxiv.org/abs/1904.07204
I'll be speaking on this at ICLR tomorrow. -
Suchi Saria
@suchisaria
Looking forward to visiting as well. Big month of travel coming up with ICLR, Mayo, ICML, Weizmann, ASTRO, and the ACM/IMS talks. I’ll be speaking about reliable ML, and ML to drive care.
Rahul Kashyap, MD, MBA 🇺🇸🇮🇳
@Dr_SRKashyap
Prof @suchisaria, Looking forward for your #keynotespeech at @MayoClinic #ai Symposium this month at #rochestermn
-
Suchi Saria
@suchisaria
Made my day! <3
Pooja
@algurlrhythm
Some of my favorite people
Professor @AnimaAnandkumar Professor @suchisaria
@vaidehijoshi @AnjanaVakil
@m0neysha @binitamshah
@Srish_Aka_Tux
Thank you for all the amazing work you do! Much ❤ to y'all! -
Suchi Saria
@suchisaria
NYU won the #icca a capella championships! Going to have to scout out their local performances. The teams this year were all so good.
-
Suchi Saria reposted
@emilymbender.bsky.social
@emilymbender
I've seen several different #nlproc folks suggesting today that it would fun/interesting/worthwhile to use BERT or GPT-2 to fill in the redacted bits of the Mueller report. A short thread on why this is a terrible idea /1
-
Suchi Saria
@suchisaria
I'm constantly surprised by how new grads choose their next jobs. I recommend reading this short thread by PG.
Paul Graham
@paulg
Programmers getting their first jobs after school don't do nearly enough research about where would be best to work. Largely, I think, because they're in denial about the inconvenient truth that research is necessary.
-
Suchi Saria reposted
Robert Nelsen
@rtnarch
Insitro and Gilead ink promising deal in NASH. Driven by machine learning. Hint of the future... endpts.com/with-150m-and-a-new-nash-discovery-deal-with-g…
-
Suchi Saria reposted
Association for Computing Machinery
@TheOfficialACM
Don't miss "Deep Learning, Reinforcement Learning, and Role of Methods in Data Science" with panelists Shirley Ho, Sham Kakade (@ShamKakade6), Suchi Saria (@suchisaria) and Manuela Veloso, at the ACM/IMS #datascience summit.
6/15, #sanfrancisco: www.acm.org/data-science-summit -
Suchi Saria reposted
Sarah Zweifach MD MBA
@sarah_zweifach
Great day at @HopkinsMedicine with the @bayesianhealth team 😍 @suchisaria @pschulam #ai #ml #sepsis #digitalhealth
-
Suchi Saria reposted
Adarsh Subbaswamy
@_asubbaswamy
Good morning Okinawa! On the first day of #aistats2019 come to poster 97 to hear me talk about @pschulam and @suchisaria ‘s work on assessing the reliability of model predictions at test time. arxiv.org/pdf/1901.00403.pdf 1/
-
Suchi Saria reposted
Iiris Sundin
@iirisSun
"Active Learning for Decision-Making from Imbalanced Observational Data", preprint with @pschulam, @eerosiivola, @avehtari, @suchisaria and @samikaski
arxiv.org/abs/1904.05268
@FCAI_Fi #fcai #aaltopml -
Suchi Saria reposted
Dana-Farber Data Science
@dfcidatascience
@suchisaria explains how novel #machinelearning techniques can develop more precise measures of disease progression and optimize treatment management in Parkinson’s disease. #zelen2019
-
Suchi Saria reposted
Rafael Irizarry
@rafalab
Looking forward to our Data Science in Biomedical Research Conference next Friday 4/5. This year's theme is "Learning from other Disciplines". Speakers are @atulbutte @_brohrer_ @cocteau @ta_broderick @StatModeling @suchisaria @Caroline_OF_B @jtleek. Available seats will be FCFS.
-
Suchi Saria
@suchisaria
Two observations about the Turing awardees—they worked on ideas that they believed in for a very long time even when these weren’t trendy. And, they were all “immigrants”.
Adding my congrats to @ylecun @geoffreyhinton & Yoshua for this special prize
www.nytimes.com/2019/03/27/technology/turing-award-ai.html -
Suchi Saria reposted
Eran Segal
@segal_eran
Few places left at our Big Data in Healthcare symposium w @NatureMedicine this May!
Register: www.weizmann.ac.il/conferences/BD2019/
Amazing speakers: @JeffDean @suchisaria @zakkohane @ShalitUri @noemieelhadad @BeEngelhardt @23andMe @_MiguelHernan @RachelRamoni @drnigam @RanBalicer -
Suchi Saria
@suchisaria
Last known photo of me with my two month old iPhone before it flew out of my hand at 2000ft into the Great iCloud Beyond...
Be safe out there folks. #mars2019 -
Suchi Saria reposted
Talking Machines
@TlkngMchns
In this episode, @lawrennd shares a favorite bit of research he came across in a conference. We discuss @suchisaria’s work on stable link in graphs. Tune in here:
www.thetalkingmachines.com/episodes/bezos-paradox-and-mac… -
Suchi Saria reposted
Uri Shalit
@ShalitUri
I highly recommend this paper. It gives a very lucid example of how identification works (and fails) in simple causal models; it also has a surprising result showing that recovery of latent factors leads to violations of positivity in multi-cause models.
Alexander D'Amour (@alexdamour.bsky.social)
@alexdamour
(Better late than never) A thread announcing my AISTATS paper: On Multi-Cause Causal Inference with Unobserved Confounding: Counterexamples, Impossibility, and Alternatives. arxiv.org/abs/1902.10286.
👇 -
Suchi Saria
@suchisaria
A thoughtful post by John Langford on reproducibility and code-sharing:
hunch.net?p=11377237 -
Suchi Saria
@suchisaria
Haha.
Karandeep Singh
@kdpsinghlab
My favorite type of inference is casual inference.
Don’t have a DAG? Don’t know what a DAG is? No worries. Just keepin’ the inference casual today. -
Suchi Saria
@suchisaria
Another terrific podcast: this one’s w/ George Church & Jorge Conde.
Highlights for me: If whole genome sequencing with CRISPR can prevent diseases, why has it not taken off? And, what are the criteria for entering the Church lab?
They’re both great communicators too.
Jorge Conde
@JorgeCondeBio
What’s in the Water at the George Church Lab? A conversation with my friend and co-founder @GeoChurch re: the past, present and future of #genomics, #crispr + #synbio — and how his lab’s radical, transformative technologies become startups 🧙♂️🧬 a16z.com/2019/02/26/george-church-lab-genome-sequencing-c…
-
Suchi Saria reposted
Johns Hopkins Bloomberg School of Public Health
@JohnsHopkinsSPH
In an article on three emerging technologies that could help physicians deal with electronic health records (EHR) systems, @JHSPH_HPM professor @withyouDrWu shouts out a machine-learning-based system developed by @SuchiSaria and used at @HopkinsMedicine. www.forbes.com/sites/insights-intelai/2019/02/11/solving-…
-
Suchi Saria reposted
Dr. Faisal Rahman
@frahman480
www.forbes.com/sites/insights-intelai/2019/02/11/solving-… @suchisaria new tools using #machinelearning to help use 21st century tools in #healthcare practice & reducing #burnout and reduce #mortality @Forbes
-
Suchi Saria
@suchisaria
Anybody else seeing these issues?
1) google scholar won't show some papers even though they've been on ArXiv for 45+ days
2) the citation on scholar references ArXiv when paper was published elsewhere >6mo ago
Many use scholar 4 citations so it's troubling. Fixes? @earnmyturns -
Suchi Saria
@suchisaria
NeurIPS needs 4500 (qualified) reviewers this year! Looks infeasible.
@hugo_larochelle it may be worth trying a process w/ two rounds. The first round—like top journals—rules out papers unlikely to score more than 3. The papers that make the cut get multiple reviewers.
Hugo Larochelle
@hugo_larochelle
Along with my NeurIPS 2019 co-Program Chairs (Alina Beygelzimer, Emily Fox, Florence d’Alché-Buc), we'll be sharing updates on preparations for the NeurIPS deadline. Here's our first blog post:
medium.com/@NeurIPSConf/starting-preparations-for-the-neu…
Enjoy, and stay tuned for more! -
Suchi Saria reposted
Stephanie Homan
@SteHoman
Great lecture on how to make #ml safe & robust by @suchisaria - part of the NYC data science seminar series @nyuniversity
-
Suchi Saria
@suchisaria
Thanks, so kind! Was very fun to put together. For all who’ve asked, we will post slides soon.
-
Suchi Saria
@suchisaria
How practitioners outside the field view #ml/AI vs those who build these methods!
--
So important to understand, identify, & close reliability gaps as ML gets used in the real world. Looking forward to tonight's tutorial and discussion on "Reliable ML". #fat2019 -
Suchi Saria
@suchisaria
100s of papers later, nearly done w/ tutorial.
In building #ml methods (& more recently live software) 4 safety-critical (SC) domains like healthcare—where cost of screwing up is high—have felt we need to increase emphasis on reliability, much like other SC disciplines. #fat2019
Suchi Saria
@suchisaria
For a tutorial at FAT*, I’m covering ideas from Reliability Engineering for building failure-proof ML systems. This is separate from fairness & ethics.
We’re covering work on reliability (proactive training, monitoring), safe AI, data shift. Your fav papers on the topic? Tx! -
Suchi Saria reposted
jake hofman
@jakehofman
Exited to have @suchisaria speaking about reliable machine learning for healthcare at next Monday's New York Data Science Seminar Series.
Register here: www.eventbrite.com/e/new-york-data-science-seminar-series… -
Suchi Saria
@suchisaria
Folks, How are tutorials cited?
I see the following as e.g.:
libguides.library.kent.edu/SPSS/citation
or
www.uvm.edu/~ddesanto/ElectroAPA2011.pdf
Meetings don't always maintain a permanent link: OK to use personal url links? does ArXiv take tutorial slides? Also, should title or publisher say tutorial? Tx! -
Suchi Saria
@suchisaria
Taking a break and...
Amazon buys Mindy Kaling’s “Late Night” for a record breaking deal at Sundance.
www.vulture.com/2019/01/mindy-kaling-late-night-sundance-…
Can’t wait to see this! (...alright, back to tutorial writing). -
Suchi Saria
@suchisaria
For a tutorial at FAT*, I’m covering ideas from Reliability Engineering for building failure-proof ML systems. This is separate from fairness & ethics.
We’re covering work on reliability (proactive training, monitoring), safe AI, data shift. Your fav papers on the topic? Tx!
Suchi Saria
@suchisaria
To clarify, esp. looking for papers that identify model agnostic failures (I’ve already inc. adversarial training, robustness, proactive methods). Also, any important but not yet popular ideas that would benefit from exposure. I’m anxious about missing good work.
DM is fine too. -
Suchi Saria reposted
Jen Heemstra
@jenheemstra
This morning, I received an email from my PhD advisor that was so encouraging and kind, it made me tear up a bit.
I finished my PhD almost 14 years ago.
Great mentors are mentors for life. If you’re just starting out, choose your PhD advisor wisely. You deserve a great mentor. -
Suchi Saria
@suchisaria
AI researchers seeking to understand humor —> start with headlines from the onion & crowdsource “unfun” ways to reword. Then learn a model from it. You can help by playing: unfun.me
I’ve secretly wanted to do standup; if this project succeeds, there’s hope!
Eric Horvitz
@erichorvitz
Computer science reveals why this headline is not funny @EPFL_en @MSFTResearch #humor #computational humor www.technologyreview.com/s/612776/computer-science-reveal…
-
Suchi Saria
@suchisaria
Superb read: @MoAlQuraishi on Deepmind’s AlphaFold & protein structure prediction at CASP13
Did DM contribute? Y
Was academia far off? N
Why didn’t pharma succeed? Don’t know
Were similar advances previously made but not advertised? Y moalquraishi.wordpress.com/2018/12/09/alphafold-casp13-wh… -
Suchi Saria
@suchisaria
Wow, painful to see individuals stuck with such bills.
Is there a consumer health startup addressing this problem?—when caught in an accident, the app tells you which facilities are in-network & it gives you option to call an ambulance to get you there.
www.vox.com/policy-and-politics/2019/1/7/18137967/er-bill… -
Suchi Saria
@suchisaria
In 2014, I’d written this editorial (citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.871.627…) to describe why #ai/ML researchers should increase focus in healthcare.
@EricTopol ‘s review below is a terrific summary of progress to date & gaps. Exciting to see the staggering pace at which the field has grown!
Eric Topol
@EricTopol
High-Performance Medicine: my @NatureMedicine review of #ai and where we're headed www.nature.com/articles/s41591-018-0300-7
(open-access for 30 days) -
Suchi Saria reposted
Thomas G. Dietterich
@tdietterich
Nice thread. @suchisaria is doing some of the best work on reliable, robust, and auditable machine learning for high-risk applications. Lots of good stuff to read here!
Suchi Saria
@suchisaria
Just resurfacing frm 2wks off over the holidays. Visited South Africa—saw gorgeous 🦒🐘🦁s, birds, seaside towns, & spent time thinking! Highlights from 2018(warning: long thread). It's been a hard yr so both bragging & celebrating. Looking fwd to all that’s in store for 2019! 1/
-
Suchi Saria
@suchisaria
Just resurfacing frm 2wks off over the holidays. Visited South Africa—saw gorgeous 🦒🐘🦁s, birds, seaside towns, & spent time thinking! Highlights from 2018(warning: long thread). It's been a hard yr so both bragging & celebrating. Looking fwd to all that’s in store for 2019! 1/
Suchi Saria
@suchisaria
First, on the research front, we made headway in multiple open directions.
As ML is being deployed in domains like healthcare, education, & recruiting, it’s critical we understand scenarios in which model outputs may be unreliable during deployment. 2/Post not available here, view on X
Suchi Saria
@suchisaria
Received a named chair--an honor typically given to v. senior faculty. What’s extra special about this is that these chairs were given to scale the very types of ideas my lab seeks to advance, ideas that were entirely new & seen w/ skepticism by many 5yrs ago. 20/
Suchi Saria
@suchisaria
Reminds me of this time in college that one of my mentors told me that I should consider a different career instead of a CS PhD. I was both confused and shattered at the time. But glad that I didn’t listen. 21/
Suchi Saria
@suchisaria
This year will likely be a year of many firsts again. The first time is often the hardest especially if you’re attempting to do something uncharacteristic. Let the naysayers not discourage.
Closing on a high note, here is a picture I took yesterday before taking off. To 2019! -
Suchi Saria
@suchisaria
Good morning George. Here for DALI!Up early prepping for a talk at the fairness, accountability, transparency session. Presenting on some of our latest work on increasingly model reliability as we move from train to deployment: arxiv.org/pdf/1812.04597.pdf & arxiv.org/pdf/1808.03253.pdf
-
Suchi Saria reposted
Eric Topol
@EricTopol
Many have questioned the benefits of #digitalhealth interventions. So I put together this Table of 10 recent RCTs and 1 impressive population study (for asthma) that cover many common medical conditions.
-
Suchi Saria
@suchisaria
Healthcare tweeps who follow me, this is worth the read.
Andy Slavitt
@ASlavitt
Medicare changed how it paid care to reward quality & improve efficiency in a pilot. 44 groups of physicians & hospitals saved the government $167 million in one year. 👇Which means they saved a multiple of that.
To me what’s important... 2/
www.cms.gov/newsroom/press-releases/cms-finalizes-pathway… -
Suchi Saria
@suchisaria
This.
Kudos to @DavidDuvenaud for not overselling: “The paper offers a proof of concept... Like any initial technique proposed in the field, it still needs to fleshed out, experimented on, and improved...”
Funnily, the editor remarks “They still need to work on their branding.” -
Suchi Saria reposted
Brian Li Han Wong
@brianwong_
Full house for the final #youandai lecture feat. @ProfBrianCox, Dr. Vivienne Ming (@neuraltheory/@SocosLabs), Prof. @SuchiSaria (@JohnsHopkins) & Prof. Peter Donnelly (@genemodeller, @UniofOxford @Genomicsplc) | @BarbicanCentre @RoyalSociety @Google @DeepMindAI @DeepMind_Health
-
Suchi Saria reposted
Alex Matthews-King
@AMatthewsKing
Acceptability of AI in medicine and diagnosis comes down to trust, whether that's a human or computer, says Prof @SuchiSaria at @royalsociety #youandai lecture with @ProfBrianCox, Dr. Vivienne Ming @neuraltheory & Prof Peter Donnelly @genemodeller
-
Suchi Saria
@suchisaria
What are some good seminal references for dataset shift? The papers don’t have to explicitly use this term. Alternatively, this idea is so fundamental, should it be getting any one citation? I’ve seen citations all over the place and want to fix. #neurips18 #neurips2018
Thanks! -
Suchi Saria reposted
Genomics
@Genomicsltd
Intrigued about #ai and its role in the next decade? This fantastic @royalsociety event hosted by @ProfBrianCox with @neuraltheory @suchisaria and our own CEO @genemodeller is sold out but you can watch via www.youtube.com/user/RoyalSociety tomorrow from 19:30 GMT royalsociety.org/science-events-and-lectures/2018/12/you-…
-
Suchi Saria
@suchisaria
Strongly endorse this thread.
-
Suchi Saria reposted
David Manheim
@davidmanheim
It's nice to see @pschulam and @suchisaria's work addressing one of the fundamental problems with using most machine learning for decision support - the lack of counterfactual reasoning.
Eric Novik
@ericnovik
Reliable Decision Support using Counterfactual Models by @pschulam and @suchisaria; arxiv.org/abs/1703.10651.
-
Suchi Saria
@suchisaria
The slides were a group effort with @pschulam and @_asubbaswamy Very sad to not be able to present; it has been years since I’ve spoken at ML4H so was looking forward to it. I’m happy that @pschulam is breaking a leg.
-
Suchi Saria reposted
Martin Mundt
@mundt_martin
Absolutely enjoyed the ml-critique-correct.github.io #neurips2018 workshop. As a PhD candidate exposed to current noise & hype in ML, it was incredibly encouraging to hear the voices of reason of @hannawallach @RandomlyWalking @suchisaria @FinaleDoshi @zacharylipton & Rich Caruana. 1/2
-
Suchi Saria
@suchisaria
Giving a talk shortly on “Knowing when not to trust: Reliable Prediction by Leveraging Causal Mechanisms” at the CRACT #neurips2018 workshop. Summarizing takeaways from the last few years of work w/ @_asubbaswamy and @pschulam Join!
Room 511b: ml-critique-correct.github.io
Tom Rainforth
@tom_rainforth
Looking forward to the CRACT (ml-critique-correct.github.io) panel session this Friday at #neurips2018 with @zacharylipton @FinaleDoshi @suchisaria @RandomlyWalking @hannawallach and Rich Caruana. We are soliciting questions for the panel in advance here docs.google.com/forms/d/e/1FAIpQLSeb71dtnA6gzLBpss8r-Ft4_…
-
Suchi Saria reposted
NYU Grossman School of Medicine
@nyugrossman
We've released the largest-ever open-source dataset to speed up MRIs using #ai. Read how we've collaborated with @facebookai to provide researchers access to MR imaging and improve patient care worldwide nyulangone.org/news/nyu-school-medicine-releases-largest-…
-
Suchi Saria
@suchisaria
Spent nearly 9 months working w/ a fantastic team on this special issue to help @PLOSMedicine get set up w/ publishing clinical ML articles.
Read the issue articles here: collections.plos.org/mlforhealth
My editorial w/ @atulbutte & @DrAzizSheikh below.
Hope you enjoy it!
Speaking Of Medicine
@SOMedicine
.@PLOSMedicine #machinelearning Special Issue Guest Eds @suchisaria, @atulbutte & @DrAzizSheikh cut through the hyperbole with an accessible and accurate portrayal of the forefront of machine learning in clinical translation
blogs.plos.org/speakingofmedicine/2018/11/28/better-medic… -
Suchi Saria
@suchisaria
This is such a creative and nice use of existing data resources to create meaningful impact.
Matt Beane
@mattbeane
AI platform finds accident predictors in Waze data, gives drivers real-time feedback, 90% of drivers slow down, accidents down nearly 20%. Seems like benefit might extend to non-Waze users too, but they don’t say... www.zdnet.com/article/waze-fed-ai-platform-helps-las-vega…
-
Suchi Saria reposted
Eric Topol
@EricTopol
Emergency room doctors have a tough time reading 12-lead electrocardiograms. So why not get a deep neural network to help them? They did, and it outdid the old, rules-based, error-laden ECG algorithm we use all the time
www.sciencedirect.com/science/article/pii/S0022073618302292 #ai @CardioLogs @UMNHealth -
Suchi Saria
@suchisaria
Most talks focus on a discovery. Wish there were more talks/papers on *how* the discovery came to be, ie, in chronological order describe the intuitions they had, what they chose to pursue, & see the chaos of the scientific discovery process in its full glory! Books that do this?
-
Suchi Saria
@suchisaria
Wow! @MikeBloomberg to donate $1.8B to enable Hopkins to offer need blind admissions forever—no more student loans—& access to “high-quality education beyond the overly rich, white communities that have traditionally enjoyed it.” <3
#giftthatkeepsongiving
www.nytimes.com/2018/11/18/opinion/bloomberg-college-dona… -
Suchi Saria
@suchisaria
Having sifted through over 50+ papers this year (guest editing a special issue for @PLOSMedicine & reading/reviewing), here are common issues I see (thread below):
Gary Collins
@GSCollins
Having spent most of today ploughing through articles on AI and machine learning for couple of papers I’m writing - the blind faith that some place on these methods is really worrying. Weakly evaluated black box data-driven medicine concerns me.
Post not available here, view on X
Suchi Saria
@suchisaria
5/ Papers have word limits so only possible to do so much. Editors want papers w/ end-to-end solutions otherwise considered not interesting. Unfortunately, papers that incrementally advance a piece of the pipeline but do a deep dive are more valuable in practice. Other ideas?
Suchi Saria
@suchisaria
Fortunately, this field is moving fast w/ many talented people getting engaged so I expect to see a lot of progress in the next decade.
-
Suchi Saria
@suchisaria
Congrats @JeffDean and @GoogleBrain This hire should be transformative in coordinating the various teams and setting a unified vision. Looking forward to seeing the great things that are to come from it.
Christina Farr
@chrissyfarr
Big news:
Google has a new head of health to define the strategy and coordinate across its various business units from Nest to Google Brain. We hear it's Geisinger Health's David Feinberg and he'll report to AI chief Jeff Dean.
www.cnbc.com/2018/11/08/google-hires-geisinger-ceo-david-… -
Suchi Saria reposted
Eric Topol
@EricTopol
Using targeted and timed neuro-stimulation to enable walking after (even years later) spinal cord paralysis, representing exceptional work @EPFL @EPFL_en @Nature @NatureNeuro @gcourtine and colleagues
www.nature.com/articles/s41586-018-0649-2
www.nature.com/articles/s41593-018-0262-6 -
Suchi Saria
@suchisaria
Friends, I’m looking for book recommendations that explore the concept of trust—what builds it, what ruins it, how to handle failures gracefully, how to communicate etc. I’m looking to understand it broadly so refs from any point of view—social, technical—are welcome. Thanks!
-
Suchi Saria
@suchisaria
Spent the day w/ Rory Staunton’s & Isabelle Fow’s parents who lost their children to sepsis. Filled w/ grief, love, joy, and perspective, and feeling ever so motivated. Grateful to be able to contribute to the @rorystauntonfdn mission. Much <3
en.m.wikipedia.org/wiki/Rory_Staunton -
Suchi Saria
@suchisaria
I suspect a lot of those who disagreed did so because they were concerned about what it does to continuity of the proceedings. Maybe change the acronym (ISNIPS-international symposium for neural information processing systems) & the proceedings remain as such. @hannawallach
Nando de Freitas
@NandoDF
I fully agree. Any other name that doesn’t make folks uncomfortable would be better #protestnips @NipsConference It’s a small kind thing to do to make a lot of people feel more respected and included. We all want this.
-
Suchi Saria
@suchisaria
Sepsis is the 3rd leading cause of death. 6 years of hard work studying how sepsis can be detected early and prevented by embedding algorithms within physician workflows. Eliza Strickland @newsbeagle writes about our work.
Eliza Strickland
@newsbeagle
#ai to identify #sepsis cases in hospitals: "First real example of AI being integrated into hospital operations, with data flowing from electronic medical records and alerts being incorporated into physicians’ workflows." spectrum.ieee.org/biomedical/diagnostics/hospitals-roll-o…
-
Suchi Saria
@suchisaria
Perhaps the best summary I’ve heard of what a truly inclusive team looks like.
In academia, we have gotten better at getting more diverse perspectives in the door but we aren’t dancing yet. The latter issue is pervasive all the way from student to the senior leadership level.
Rock Health
@Rock_Health
“Diversity is inviting someone to the party. Inclusion is when you dance. And dancing is a back and forth—sometimes you lead, sometimes I lead.” -@DrIvorHorn @Accolade #rhsum18
-
Suchi Saria reposted
Rock Health
@Rock_Health
Silicon Valley has failed us in behavioral health by focusing on solutions for people with insurance that can pay on top of their plan. That’s not the biggest need. -@rich_roth
Focusing on Medicaid makes the biggest impact in healthcare -@leah_e_sparks #rhsum18 -
Suchi Saria reposted
Rock Health
@Rock_Health
For entrepreneurs and the private sector on policy for innovation: “There’s a lot of opportunity to inform and course correct. But Washington can’t do it on its own.” @KBDeSalvo with @AledadeACO’s @Farzad_MD #rhsum18
-
Suchi Saria reposted
Rock Health
@Rock_Health
“If we focus on the workflow instead of replacing doctors, healthcare would be better” @SameerBadlaniMD #rhsum18
-
Suchi Saria
@suchisaria
This is a momentous shift: I love that cardiologists are pushing to engage in healthtech & advocating 4 themselves as great hires.
I’ve worked w/ collaborators from many clinical areas: the best dislike hierarchy, are humble, curious, eager to learn, & bring a LOT to the team. -
Suchi Saria
@suchisaria
I wish there was a VR company that let me go on runs anywhere in the world. Coupled with a treadmill it would simulate landscapes. Better, an AR company that lets you take bootcamp & other classes with participants, say on a beach! Does anything like it exist? How far off are we?
-
Suchi Saria
@suchisaria
<3 thank you!
Retweet-is-a-brag-but-this-is-special.
Harlan Krumholz
@hmkyale
Listening to @suchisaria @HopkinsMedicine: she is an outstanding #datascientist truly bridging to clinical medicine. Has done excellent work with sepsis detection and more #bayesian and probabilistic modeling approaches. We need more of people like her. A gem.
-
Suchi Saria
@suchisaria
This is a remarkable project.
euan ashley
@euanashley
New from the Undiagnosed Diseases Network! In 20 months, 1519 referred, 601 accepted. Diagnoses made in 35%. Almost 80% of diagnoses led to a change in medical management. 31 new syndromes. Healthcare costs after diagnosis: 6% of those before. www.nejm.org/doi/full/10.1056/NEJMoa1714458
-
Suchi Saria
@suchisaria
Seeking a postdoctoral fellow interested in augmented decision-making i.e., exploring principles for building #ai reasoning algorithms that support human experts in decision-making when safety is critical. You will be joining an amazing interdisciplinary team funded by #nsf
-
Suchi Saria
@suchisaria
Amazing news: @deborahestrin recognized as a MacArthur Fellow (also called the genius grant). Congrats Dr. Deborah Estrin! You’ve always been a fearless leader for the CS community and a wonderful source of inspiration.
www.macfound.org/fellows/1009/ -
Suchi Saria
@suchisaria
Thank you! My first time at #idweek2018 Really enjoyed the speakers, panelists and the audience questions in our session. Looking forward to the rest of it and visiting again.
@HopkinsMedicine @HopkinsEngineer -
Suchi Saria
@suchisaria
If you’re interested in AI research, this is a must read. It’s a thoughtful commentary explaining recent progress.
“Recent developments tell us more about the problems tackled & the structure of our world than about neural networks per se”
m-cacm.acm.org/magazines/2018/10/231373-human-level-intel… -
Suchi Saria
@suchisaria
Just presented to the National Academy of Sciences (NAS) Health Services Board. David Blumenthal & Carolyn Clancy are legends & heroes of mine so it was extra special leading a discussion for them on #ai, where it’s headed, & what it can & can’t do. Questions I got asked 1/
Suchi Saria
@suchisaria
What is AI/ML/DL? Can you show us working examples w/ measured benefit?
What will it take to scale AI in healthcare?
Does it have to learn over time?
Some surprising ??s: what is the diff b/w robotics & AI? Does all robotics use AI?
Will it add to cost?
Great discussion. -
Suchi Saria
@suchisaria
IBM claims new general purpose tool to detect bias in arbitrary models.
R they Watsoning again?
The right measures of bias varies by domain & remains an active area of research. At best, we can check for specific types of bias w/in individual domains
techcrunch.com/2018/09/19/ibm-launches-cloud-tool-to-dete… -
Suchi Saria
@suchisaria
#ml folks interested in causality, this is a fantastic read.
There is a tendency in our community to focus on step 6 w/ a complex deep model w/o understanding of 1-5. Clear specification of the question & assumptions abt the DGP r key.
www.ncbi.nlm.nih.gov/pmc/articles/PMC4077670/
via @yudapearl -
Suchi Saria
@suchisaria
This is looking to be a fantastic #nips2018 workshop with five diff tracks including “Bad Practices”, “Flawed Intuitions”, and “Negative Results”, and “Open Problems”. Submit! A subset will get a slot to register for NIPS.
(I’ll be speaking on bad practices & open problems.)
Tom Rainforth
@tom_rainforth
Along with my co-organizers Matt Kusner, Ben Bloem-Reddy, Brooks Paige, Rich Caruana, and @yeewhye, I'm excited to announce the first #nips workshop on "Critiquing and Correcting Trends in Machine Learning"
CFP: ml-critique-correct.github.io
Submit by Oct 23rd: easychair.org/conferences/?conf=cract2018 -
Suchi Saria
@suchisaria
I advocated for changes to training so that (1) we are seeing more physician leaders who can be co-conspirators in innovation, & (2) courses to bridge gap b/w current medical training & a future w/ health care delivery that heavily leverages AI/data-intensive tools.
Other ideas?
Suchi Saria
@suchisaria
I spent a day w/ leadership at the @AmerMedicalAssn this week to discuss role that AI/ML was going to play and what they can do to prepare!
They asked what we suggest changing with where things currently stand—got me thinking... -
Suchi Saria
@suchisaria
I spent a day w/ leadership at the @AmerMedicalAssn this week to discuss role that AI/ML was going to play and what they can do to prepare!
They asked what we suggest changing with where things currently stand—got me thinking... -
Suchi Saria reposted
Hagai Rossman
@H_Rossman
Great slides on ML and Causal Inference from @suchisaria. Highlights practical methods and filled with lots of great references. Thank you for sharing with the community
Suchi Saria
@suchisaria
Slides from MLSS@Madrid: www.dropbox.com/s/m233izdg1mzbee2/MLSS-18_final.pdf?dl=0
Gave lectures on “ML+Causal Inference for Reliable Decision Support”
This group was super curious & engaging—was easy to stay excited even w/ only 5 hrs of sleep in 2 days! Thank u Drs Hernández-Lobato 4 hosting! -
Suchi Saria
@suchisaria
Slides from MLSS@Madrid: www.dropbox.com/s/m233izdg1mzbee2/MLSS-18_final.pdf?dl=0
Gave lectures on “ML+Causal Inference for Reliable Decision Support”
This group was super curious & engaging—was easy to stay excited even w/ only 5 hrs of sleep in 2 days! Thank u Drs Hernández-Lobato 4 hosting! -
Suchi Saria reposted
Nature Biotechnology
@NatureBiotech
Be a part of the conversation on #bigdata and Cancer #precisionmedicine with @ThomasFuchsAI @corrie_painter and @suchisaria. Check out the meeting program go.nature.com/2KRWrYa #bigdatacpm18
-
Suchi Saria reposted
Vijay Pande
@vijaypande
From all my experience, @suchisaria is one of the very most impressive biomedical scientists I know. Amazingly creative and able to lead others to create massive impact. #thiswomanrocks
Esther Choo, MD MPH
@choo_ek
1. Pick a female colleague
2. Describe why she’s awesome
3. Use the hashtag #thiswomanrocks
Let’s hear about a ton of cool women! -
Suchi Saria
@suchisaria
. @DaphneKoller has inspired & trained many (fe)male scientists & turned them into impact powerhouses. She is committed to excellence (not exaggerating), is brilliant, cares deeply about everything she takes on, & leads by example. #thiswomanrocks
(DK, sorry 2 embarrass u.)
Esther Choo, MD MPH
@choo_ek
1. Pick a female colleague
2. Describe why she’s awesome
3. Use the hashtag #thiswomanrocks
Let’s hear about a ton of cool women! -
Suchi Saria reposted
National Academy of Medicine
@theNAMedicine
Suchi Saria is one of our new Emerging Leaders! As an Assistant Professor @JohnsHopkins, her research focuses on developing next generation diagnostic and treatment planning tools that leverage statistical methods to individualize care. Welcome, @suchisaria!
-
Suchi Saria reposted
Pieter Abbeel
@pabbeel
Great reading list on Safety and Control for AGI:
inst.eecs.berkeley.edu/~cs294-149/fa18/
taught by Andrew Critch and Stuart Russell from @berkeley_ai -
Suchi Saria
@suchisaria
The video below is delightful.
Who wants to play a bet for which school is going to go next and how long till we hear the next such announcement?
Huge respect for this program’s donors setting such a nice precedent.
NYU Grossman School of Medicine
@nyugrossman
Watch their reactions as our #medstudents learn they will be receiving full-tuition scholarships! #whitecoatceremony
Suchi Saria
@suchisaria
Curious which of my three “alma maters”—Stanford, Harvard and Hopkins—are going to go first? Time for some healthy competition!
@deniswirtz @Stanford @JohnsHopkins -
Suchi Saria
@suchisaria
Excited about this line of work w/ @_asubbaswamy & @pschulam tackling issues that hurt reliability as we deploy ML models.
This UAI’18 paper tackles selection bias & domain dependent confounding & proposes a way 2 identify stable relationships that generalize from train to test.
Adarsh Subbaswamy
@_asubbaswamy
Our (w/ @suchisaria) newest paper on improving reliability when deploying machine learning models for decision making. 1/
arxiv.org/abs/1808.03253 -
Suchi Saria
@suchisaria
Very honored to be included. And especially excited to see the NAM integrate individuals w/ backgrounds as diverse as mine. Looking forward to working with this incredible group over the next few years.
@HopkinsMedicine @HopkinsEngineer
@ThatClaire @JGreenGroup @DrDeidraCrews
National Academy of Medicine
@theNAMedicine
We are proud to announce the inaugural class of Emerging Leaders in Health and Medicine! These early- and mid-career professionals are working to better the future of their respective fields. Meet the leaders here: nam.edu/our-work/health-policy-fellowships-and-leadership…
-
Suchi Saria reposted
Nicholas A. Peppas
@NPeppas
Jordan Green JHU selected an “Emerging Leader in Health and Medicine” by @theNAMedicine.
A prestigious program recogn. emerging leaders in Medicine. The inaugural Emerging Leaders Forum will be in Washington, DC on April 11-12, 2019. nam.edu/programs/emerging-leaders-forum/ Congrats @JGreenGroup -
Suchi Saria reposted
Steven Salzberg 💙💛
@StevenSalzberg1
Impressive work by @mike_schatz digging into the complex details of cancer genome rearrangements using long-read sequencing technology.
Michael Schatz
@mike_schatz
Research Highlight in @nature on our efforts to sequence breast cancer with long reads: Tallying a cancer’s vast array of genetic changes
www.nature.com/articles/d41586-018-05670-4 -
Suchi Saria
@suchisaria
Looking forward to the Machine Learning + Causal Inference workshop today sites.google.com/site/faim18wscausalml/program-schedule
Don Rubin is starting us off!
I’ll be speaking on improving prediction model reliability by leveraging causal mechanisms, joint work w/ @pschulam and @_asubbaswamy #icml2018 -
Suchi Saria reposted
James Owers-Bardsley
@jamesowers
Holy. Shit. Expressive, interesting, and lengthy piano music generations. I genuinely got goosebumps listening to these. As far as I'm concerned, this is a huge step forward for creative ML. Cheers Anna Huang. Paper: docs.google.com/viewer?a=v&pid=sites&srcid=ZGVmYXVsdGRvbW… #icml2018
-
Suchi Saria
@suchisaria
I just got a CV for a postdoc who looks strong and is interested in probabilistic methods + genomics. I'm not taking someone on that topic right now but if you're looking for someone, ping me. I'll let them know.
-
Suchi Saria reposted
Neil Lawrence
@lawrennd
A very comprehensive paper for those into kernel methods and GPs.
Gives deep insights into the nature of our assumptions when modelling.
Thanks Motonobu Kanagawa, Philipp Hennig, @sejDino and Bharath K Sriperumbudur!
arxiv.org/pdf/1807.02582.pdf -
Suchi Saria
@suchisaria
Honored to be joining an incredible group of researchers ( nam.edu/programs/emerging-leaders-forum/ ) as part of the NAM Emerging Leaders program and bringing the “younger” generation’s perspective to it. Expect to see more on ML, AI, data science & health!
w/ @JGreenGroup @DrDeidraCrews -
Suchi Saria
@suchisaria
<3
Mount Sinai Health System
@MountSinaiNYC
An Oklahoma husband made his wife fall in love with him again after she suffered from a form of #amnesia that made her forget all their memories together. Dr. Kirk Lercher weighs in on her condition via @ABC: abcnews.com/GMA/Family/wife-amnesia-accident-couple-dates…
-
Suchi Saria
@suchisaria
Recommend.
And, “issue here is that it’s easier to come up with a method than to get it to work reliably, another seems to be a problem in academic incentives, where there’s more of a motivation to attack new problems than to get old things working.” Yup. andrewgelman.com/2018/06/27/yes-work-evaluating-variation… -
Suchi Saria reposted
Eric Topol
@EricTopol
An exceptional review of CAR-T (T-cell engineering) therapy by a couple of its pioneers
www.nejm.org/doi/full/10.1056/NEJMra1706169?query=feature… @NEJM by June and Sadelain -
Suchi Saria
@suchisaria
Terrific article by the talented @newsbeagle uncovering the issues surrounding the IBM Watson layoffs. I’m relieved that the explanations point to lack of customer obsession and not that health systems don’t care about improving care and outcomes.
spectrum.ieee.org/the-human-os/robotics/artificial-intell… -
Suchi Saria
@suchisaria
Clever use of ML to challenge a widely held belief that end-of-life spending is often “wasteful” spending.
“Death is highly unpredictable. Less than 5% of spending is accounted for by individuals with predicted mortality above 50%.”
science.sciencemag.org/content/360/6396/1462.full
congrats @oziadias -
Suchi Saria reposted
-
Suchi Saria
@suchisaria
Traveling for 2 days and left my wallet at home—yikes! Can I survive? This is going to be a true test of how digital our lives have become.
Just got Apple Pay setup. First stop: Starbucks. I have a london fog and sous vide egg bites in hand!
Suchi Saria
@suchisaria
I was worried that lunch was going to be tricky b/c most restaurants don’t accept Apple Pay. Turns out, I had to present at a meeting this afternoon and they’d generously ordered delicious lunch (with tea) for the presenters! Always nice when things line up this conveniently :-)
Suchi Saria
@suchisaria
Thought I’d have to skip dinner yesterday but my postdoc fellow offered to take me out :) That place had Apple Pay so phew I didn’t have to borrow.
Now traveling home; found a trailmix that looked great but alas...
until digging in my bag led to a $5 bill!
I think I’ll make it! -
Suchi Saria
@suchisaria
It would be great if we had funded @NIH or @AHRQ studies on this topic.
Eric Topol
@EricTopol
How much time will #ai free up for clinicians?
The @IPPR in-depth report has projections, along w/economic impact
www.ippr.org/research/publications/better-health-and-care… #oa via @SueLaceybryant @NHS_HealthEdEng -
Suchi Saria
@suchisaria
Keynote started with an intro that said Dr. Gawande and Dr. Makary (@MartyMakary) were some of the previous speakers! I had big shoes to fill.
Some tough questions about whether hospitals will be anxious to deploy such tech (but some large health systems chimed in saying no!)
Suchi Saria
@suchisaria
Listening to @NIH’s Dr. Devaney’s keynote about the #allofus 1Mi cohort project at the American Health Lawyers Association annual meeting. My keynote is next.
I’m really looking forward to hearing the questions individuals here have in mind. #ahla -
Suchi Saria reposted
Valita Fredland, JD, MA
@groovycounsel
Augmenting Clinical Intelligence using "Maching Intelligence" @suchisaria Exciting emergence of research taking advantage of "secondary sources of data" and related ethical and legal challenges. @healthlawyers #ahla18 for #healthit
-
Suchi Saria
@suchisaria
Listening to @NIH’s Dr. Devaney’s keynote about the #allofus 1Mi cohort project at the American Health Lawyers Association annual meeting. My keynote is next.
I’m really looking forward to hearing the questions individuals here have in mind. #ahla -
Suchi Saria
@suchisaria
An observation: Hopkins & Harvard have the two largest schools of public health. But among physician execs/entrepreneurs, many seem to have their roots at the MPH program @ Harvard.
Is this true or happens to be my sample set? If true, why? Curious.
Suchi Saria
@suchisaria
For those attributing to Boston ecosystem, can you elaborate how you ended up choosing to take a more entrepreneurial path. Did you:
-
Suchi Saria
@suchisaria
Very insightful read on traits associated with good, lasting relationships: www.theatlantic.com/health/archive/2014/06/happily-ever-a…
-
Suchi Saria reposted
Eric Horvitz
@erichorvitz
On understanding "failure to rescue" in hospitals leading to death www.bloomberg.com/news/articles/2018-06-22/training-algor… There's great upside to harnessing machine learning and decision making to enhance care and to build insights. @AMIAinformatics @MSFTResearch @suchisaria @compcomcon
-
Suchi Saria
@suchisaria
Exactly right!
So stoked about this choice. Feels like the entire community is routing for this combo to succeed! Congrats again @Atul_Gawande
Farzad Mostashari
@Farzad_MD
“We said at the outset that the degree of difficulty is high and success is going to require an expert’s knowledge, a beginner’s mind, and a long-term orientation,” said Jeff Bezos, founder and CEO of Amazon. “Atul embodies all three”
+1 @Atul_Gawande
www.businesswire.com/news/home/20180620005747/en/Amazon-B…. -
Suchi Saria
@suchisaria
Wow! Great choice.
www.businesswire.com/news/home/20180620005747/en/Amazon-B…. -
Suchi Saria
@suchisaria
Breaking news: The @AmerMedicalAssn releases a statement re: their stance on AI. Pleased to see them recognize value AI can provide in improving care. Need flexibility re:implementation as we learn about best ways to incorporate new learning tech.
www.ama-assn.org/ama-passes-first-policy-recommendations-… -
Suchi Saria reposted
Scientific American
@sciam
"A big application of robotics is elder care. In Japan, its becoming common for robots to take care of humans. They are designed to be attentive to peoples needs." —@pmddomingos #scienceonthehill
-
Suchi Saria reposted
Sen. Jerry McNerney
@SenMcNerney
Kicking off a great discussion about artificial intelligence and its role in the future of healthcare. #scienceonthehill
-
Suchi Saria reposted
Scientific American
@sciam
Q&A: What do you think about the quality of the data in the medical record? Listen to @suchisaria’s answer. #scienceonthehill
This media couldn't be displayed. See it on the original post. -
Suchi Saria reposted
Scientific American
@sciam
Tomorrow: Follow #scienceonthehill for live updates from our discussion at Capitol Hill on how AI and robotics will shape health care in the future. Speakers include @mdichristina, @RepMcNerney (CA-09), @suchisaria, @pmddomingos and Pamela Hepp. blogs.scientificamerican.com/at-scientific-american/scien…
-
Suchi Saria
@suchisaria
Great opportunity for rising PhD students who want exposure to diverse points of view for technical work in medicine and healthcare.
This is the first of multiple workshops that will take place over the 2018-2019 calendar year. Bookmark the cite for future announcements.
Miguel Hernán
@_MiguelHernan
Interested in #datascience for precision medicine?
The SAMSI Program on Methods for Precision Medicine will hold its first workshop on August 13-17 at North Carolina State Univ.
Students are encouraged to apply (support available!). Deadline: June 18
www.samsi.info/programs-and-activities/year-long-research… -
Suchi Saria
@suchisaria
Thoughtful and important point by @DShaywitz
(Need to shake off commitment to status quo.
Everyday that we choose to block innovation, we are implicitly saying yes to losing a few more lives that could have otherwise been saved.
And, no, not advocating 4 Theranos’ approach.)
David Shaywitz
@DShaywitz
The (wrong) lesson I fear many may conveniently take away from Theranos debacle. www.forbes.com/sites/davidshaywitz/2018/06/04/learning-th… @JohnCarreyrou
-
Suchi Saria
@suchisaria
Wow! Just got the most sophisticated spam call from the "IRS"---they identified themselves using an IRS badge #, asked me home add for verification, described missing amounts in my tax filings and said if I didn't pay, an arrest warrant would be issued!
www.irs.gov/newsroom/irs-warns-of-pervasive-telephone-scam -
Suchi Saria reposted
Reza Zadeh
@Reza_Zadeh
FDA approved marketing of an AI algorithm that detects wrist fractures from x-rays. This is the second such approval from the FDA for AI, showing signs of speeding up approvals for AI-powered diagnostics. Great step forward. www.fda.gov/NewsEvents/New
-
Suchi Saria
@suchisaria
Twitter as a medium for advertising scientific results is partly to blame for this erosion of trust.
Many scientists are taking to twitter broadcast results. 1/
Lior Pachter
@lpachter
And how can the public trust scientists if time and time again they are presented with hype instead of science? /5
Suchi Saria
@suchisaria
Scientific communication often requires subtle qualifiers & reliance on assumed knowledge to communicate. The less steeped someone is in the area, the more background explanation is needed to explain the significance of a result. Social media doesn’t facilitate such discourse. 2/
-
Suchi Saria reposted
Sebastien Bubeck
@SebastienBubeck
The list of accepted papers for COLT 2018 has been posted! I am extremely pleased with the selection. Very high quality program that showcases the diversity of modern learning theory (optimization, bandits, RL, neural networks, fairness/privacy, etc..).
www.learningtheory.org/colt2018/call.html -
Suchi Saria reposted
Farzad Mostashari
@Farzad_MD
This a thousand times
Kevin Volpp
@kevin_volpp
The double standard whereby preventive services must show positive ROI to be covered but treatments must only have benefits>risk has several reasons and deleterious consequences for #population #heath summarized in this @k_pryorMD article @NEJM @PennCHIBE www.nejm.org/doi/full/10.1056/NEJMp1716272
-
Suchi Saria reposted
francesca dominici
@francescadomin8
A VERY conservative estimate is that the Trump environmental agenda is likely to cost the lives of over 80 000 US residents per decade and lead to respiratory problems for many more than 1 million people. @harvard_data @EPA @GinaEPA @HealthyAirUK
francesca dominici
@francescadomin8
JAMA Forum: A Breath of Bad Air: Trump Environmental Agenda May Lead to 80 000 Extra Deaths per Decade newsatjama.jama.com/2018/05/10/jama-forum-a-breath-of-bad… via @newsatjama
-
Suchi Saria
@suchisaria
Aww. Thank you! That’s very kind of you to say! <3 I have to call out my amazing students who I’m super proud of and have been my partners in crime.
Fei-Fei Li
@drfeifei
Haha Suchi, and look at what you’ve accomplished and done for the world since then! Your alma mater and everyone here is very proud of you <3
This post is unavailable. See it on X.
Suchi Saria
@suchisaria
A subset of my them are here (and yes I need a better lab photo).
-
Suchi Saria
@suchisaria
Surprised to be selected to this. As I’m browsing, I see incredible people both in this year & past years’ cohorts—there are scientific leaders like the one of two CRISPR inventors and so on AND...wait for it..Priyanka Chopra!!
Thank you to my undisclosed nominators!
Young Global Leaders
@YGLvoices
Meet the 2018 @wef Young Global Leaders #ygl18. Find out who they are 👉 wef.ch/ygl18
-
Suchi Saria
@suchisaria
Neurology Today of the American Academy of Neurology covering our work in Parkinson's: “It represents the first demonstration of harnessing [ML] for measurement and monitoring of patients, which can serve to optimize treatment management,” Dr. Espay said.
journals.lww.com/neurotodayonline/Fulltext/2018/05030/Neu… -
Suchi Saria reposted
David Bindel
@DavidBindel
I was PhD admissions chair for CS at Cornell this year. e made a big step in improving diversity of the program. Let me tell you about it.
-
Suchi Saria reposted
PLOS Medicine
@PLOSMedicine
Less than 4 weeks left to submit to our Special Issue on #machinelearning in health and biomedicine!
w/ Guest Editors @atulbutte, @suchisaria & @DrAzizSheik
Deadline May 25
plos.io/machinelearningmed
#ml4health -
Suchi Saria reposted
Johns Hopkins Bloomberg School of Public Health
@JohnsHopkinsSPH
Steven Salzberg, known for his ability to tackle the most difficult projects in genome sequencing, has been elected to the @americanacad. hub.jhu.edu/2018/04/27/steven-salzberg-american-academy-a… @JHU_BDPs
-
Suchi Saria
@suchisaria
Thank you! Here are the slides with links to all ref papers included: www.dropbox.com/s/6slyz4g7f5x5g63/2018-ICLR-Saria.pdf?dl=0
#iclr2018
Thang Luong
@lmthang
Nice #iclr2018 talk by @suchisaria on remote monitoring of Parkinson disease using existing mobile sensors & reliable risk detection using counterfactuals www.facebook.com/iclr.cc/videos/2124298770932511/ (related papers arxiv.org/pdf/1601.00960.pdf & arxiv.org/pdf/1703.10651.pdf)
-
Suchi Saria reposted
Jihoon Jeong
@hiconcep
Today’s #iclr2018 opening invited lecure by @suchisaria “Individualizing Healthcare with Machine Learning”. #ml is becoming common tool at the hospital.
-
Suchi Saria
@suchisaria
Finally made it! #iclr2018
Looking forward to seeing everyone and giving my invited talk bright and early at 9am! iclr.cc/Conferences/2018/Schedule?type=Invited Talk -
Suchi Saria reposted
Lior Pachter
@lpachter
A beautiful essay by Michael Jordan on AI calling for us to "broaden our scope, tone down the hype and recognize the serious challenges ahead." Essential reading medium.com/@mijordan3/5e1d5812e1e7
-
Suchi Saria
@suchisaria
Terrific read.
Joe Hellerstein
@joe_hellerstein
Important #longread from my colleague Michael Jordan with a measured academic look at AI hype and computational/statistical reality and opportunity around Data, ML, and yes, AI. medium.com/@mijordan3/artificial-intelligence-the-revolut…
-
Suchi Saria reposted
adaobi
@adaobiadibe_
African women in tech, apply!
Deep Learning Indaba
@DeepIndaba
Applications to participate in the 2018 Indaba close in a few weeks. Encourage as many as you know to apply at deeplearningindaba.com/apply. We can't wait to learn from the inspirational speakers we have lined up. More details on this soon ...
-
Suchi Saria
@suchisaria
Wow! MIT Tech Review is deteriorating into a fake news site. I have seen so many what appear to be nonsensical headlines recently. I don’t see any evidence in support of this click-baity headline. www.technologyreview.com/the-download/610888/facebook-is-…
Suchi Saria
@suchisaria
To clarify, I used to be a huge fan of The Tech Review. I rarely read articles with conceptual errors. I even tried to get my non-techie family to start reading it. But in the last six months, feels like the focus has shifted towards optimizing for clicks. cc @techreview
-
Suchi Saria reposted
Thomas G. Dietterich
@tdietterich
Nice thread seeking advice on model management best practices. Many useful responses.
Caitlin Hudon
@beeonaposy
Data scientists: what's your team's approach to tracking the quality of models in production?
(How do you know if a model is decaying? How do you quality-check the data going into a model? Who builds and tracks these things?) -
Suchi Saria reposted
Anshul Kundaje
@anshulkundaje
Repeat after me: A reviewers role is POLITELY point out flaws in manuscripts and provide HELPFUL suggestions with the goal of helping authors IMPROVE their manuscript. The goal is not to find some random reason to reject the paper. And definitely not a license to be an ahole.
-
Suchi Saria reposted
Kumar Garg
@KumarAGarg
Today, I was asked for any documents that explain "how your team got so much done" in Obama White House. I sent them this -> www.mitpressjournals.org/doi/pdf/10.1162/inov_a_00253
-
Suchi Saria reposted
Christina Farr
@chrissyfarr
Seriously cool (albeit early) research.
Parkinson's patients can pickup a smartphone, download an app, try out a few simple tests like finger tapping balance and reaction time, and their doctors can better manage their meds -> www.medpagetoday.com/neurology/parkinsonsdisease/71995 via @suchisaria -
Suchi Saria
@suchisaria
Just got this gorgeous bouquet for my upcoming Malone Chair installation. Feeling the love and support from my wonderful colleagues. Thanks @JHUBME @HopkinsEngineer @JohnsHopkins for making my last 5.5 years so productive and fun!
Suchi Saria
@suchisaria
To clarify, my appointment is in CS with joints (across stats, biostats, health policy); thanks is especially owed to them.
-
Suchi Saria
@suchisaria
Amusing: Machine reading put to good use!
Quartz
@qz
Slack is developing tools to tell if someone's mansplaining
qz.com/leadership/future-of-work/1238413 -
Suchi Saria reposted
euan ashley
@euanashley
Featuring excellent work from @suchisaria www.digitaltrends.com/computing/nvidia-gtc-project-clara-…
-
Suchi Saria
@suchisaria
Terrific read especially for aspiring researchers.
Berk Ustun
@berkustun
“Grungy boots-on-ground work is how we build our intuitions about what kinds of solutions actually work vs. sounding good on paper. It is hard — though not impossible — to skip this step and still do great work.” @johnregehr
blog.regehr.org/archives/1582 -
Suchi Saria reposted
JHFastForward
@JHFastForward
Another immunometabolism deal? Johns Hopkins spinout Dracen lands $36M investment from Deerfield. endpoints.news/another-immunometabolism-deal-johns-hopkin…
-
Suchi Saria reposted
JAMA Neurology
@JAMANeuro
Researchers turn to #smartphone #app and #machinelearning to develop an objective digital biomarker that tracks real-world symptom severity in those with #parkinsondisease jamanetwork.com/journals/jamaneurology/fullarticle/10.100…
-
Suchi Saria reposted
Thomas Sutton
@thomas_thinks
Inspiring example of the good side of data benefiting people, after so much bad news about abuse of data recently.
Suchi Saria
@suchisaria
Took us 4yrs to go from concept to app to the approach & clinical validation but exciting to be the first app and now the approach in use by 10K+ across both Apple and Android.
Kudos to the team esp the inimitable @AndongZhan, amazing ugrad advisee Srihari, Chris T. & Ray D. -
Suchi Saria
@suchisaria
Thanks @EricTopol Very excited abt this work.
Imagine managing diabetes w/ 3 readings/yr! PD patients experience similar symptom variability but don’t have the tools to measure fluctuations frequently. Drug trials often fail b/c of lack of ways to quantify change in severity 1/
Eric Topol
@EricTopol
How a smartphone app + #ai can digitize the symptoms of #parkinsons disease jamanetwork.com/journals/jamaneurology/fullarticle/10.100… @JAMANeuro @AndongZhan @suchisaria @HopkinsMedicine @UofR @JohnsHopkins
Post not available here, view on X
Suchi Saria
@suchisaria
Took us 4yrs to go from concept to app to the approach & clinical validation but exciting to be the first app and now the approach in use by 10K+ across both Apple and Android.
Kudos to the team esp the inimitable @AndongZhan, amazing ugrad advisee Srihari, Chris T. & Ray D.
Suchi Saria
@suchisaria
Thanks to @MichaelJFoxOrg for supporting this work!
-
Suchi Saria reposted
Eric Topol
@EricTopol
How a smartphone app + #ai can digitize the symptoms of #parkinsons disease jamanetwork.com/journals/jamaneurology/fullarticle/10.100… @JAMANeuro @AndongZhan @suchisaria @HopkinsMedicine @UofR @JohnsHopkins
-
Suchi Saria reposted
Denis Wirtz
@deniswirtz
Johns Hopkins leads all U.S. universities in research spending for 38th year in a row hub.jhu.edu/2018/03/23/nsf-research-funding-rankings/. 2.43 billions dollars!
-
Suchi Saria
@suchisaria
Looking forward to reading this! I suspect useful for anyone who is tackling ML problems w/ societal implications.
Hanna Wallach (@hannawallach.bsky.social)
@hannawallach
More exciting news!!! My latest working paper is now online!!! 🎉🎊
Datasheets for Datasets
by @timnitGebru, @jamiemmt, @brianavecchione, @jennwvaughan, me, @haldaume3 & @katecrawford.
We hope it nudges ML & AI toward more transparency re. data!!! 📊
jamiemorgenstern.com/papers/datasheet.pdf -
Suchi Saria reposted
Andy Biotech
@AndyBiotech
Special issue of @sciencemagazine on cancer #immunotherapy this week
Immune checkpoint
science.sciencemag.org/content/359/6382/1350
CAR-T
science.sciencemag.org/content/359/6382/1361
Personalized vaccines
science.sciencemag.org/content/359/6382/1355
#microbiome
science.sciencemag.org/content/359/6382/1366 -
Suchi Saria reposted
Russ Altman
@Rbaltman
Think drug discovery is easy and linear? Not so fast….Here’s the map. ncats.nih.gov/translation/maps
-
Suchi Saria reposted
Lindy Elkins-Tanton
@ltelkins
A robot parade at #mars2018 — the round ones are autonomously following the dog. Also, a drone.
This media couldn't be displayed. See it on the original post. -
Suchi Saria reposted
Fei-Fei Li
@drfeifei
A countless number of failures... :) I still fail on a daily basis. But I’ve learned the most important thing is to get up and keep trying :)
Nataniel Ruiz
@natanielruizg
What is a bad experience in your life that changed you for the good?
-
Suchi Saria reposted
PLOS Medicine
@PLOSMedicine
Announcing a new Special Issue on #machinelearning in health and biomedicine, with Guest Editors @atulbutte, @suchisaria & @DrAzizSheikh ! #ml4health
Submit by May 25, 2018
plos.io/machinelearningmed -
Suchi Saria reposted
Abigail See
@abigail_e_see
Is structure a necessary good or a necessary evil?
The video of @ylecun and @chrmanning's discussion on Deep Learning, Structure and Innate Priors is online! Watch the entire conversation, and read a summary of it, here:
www.abigailsee.com/2018/02/21/deep-learning-structure-and…
#stanford #deeplearning #ai -
Suchi Saria reposted
NVIDIA AI
@NVIDIAAI
#ai is a powerful tool for surveying and classifying biological data. Learn from experts in the healthcare field at #gtc18 including @Google’s Mark DePristo, @SuchiSaria and more: nvda.ws/2oyRiLv
-
Suchi Saria reposted
Scott Gottlieb, MD 🇺🇸
@ScottGottliebMD
“FDA plans to use a proposed $400 million boost in federal funding to focus on a range of innovative approaches to speed the approval of new medical devices and create a new center to support digital health oversight and address cybersecurity concerns.” www.fiercehealthcare.com/regulatory/fda-budget-center-exc…
-
Suchi Saria reposted
Arunesh Mittal
@arunesh_mittal
Machine Learning Summer School 2018, Madrid website (mlss.ii.uam.es/mlss2018/ ) is finally up! Really looking forward to talks by two ML researchers who always inspire me @shakir_za and @suchisaria! #mlss2018
-
Suchi Saria
@suchisaria
Terrific read. Sums up much of what I’ve learnt this past year building ML4CDS:
“Weak human + machine + better process was superior to a strong computer alone and, more remarkably, superior to a strong human + machine + inferior process.”
via @jdlovejoy jods.mitpress.mit.edu/pub/issue3-case -
Suchi Saria
@suchisaria
For me, it was visiting the Stanford NICU in 2007: I saw fragile preemies have surprise downturns, yet many were avoidable if only docs saw it coming.
Spent 3yrs working on this w/ some wins but found scaling up invention 2 be <3 wrenching. Haven’t given up.
#whatgotmeintohc
Christina Farr
@chrissyfarr
Let's start a Friday meme...
What got you interested in health care?
For me, getting sick as a teenager with no diagnosis for a year. Got obsessed with the history of medical paternalism and what it meant to be an informed patient.
#whatgotmeintohc -
Suchi Saria reposted
Iain Buchan
@profbuchan
Looking forward to discerning the needs-based ML from the noise with @NielsPeek @EnricoCoiera @DaniCMBelg @suchisaria and colleagues at the Data, Learning and Inference 2018 meeting in April.
Iain Buchan
@profbuchan
Some models may be learned from data, others may not, others need not. ML has a long statistical history pre-dating the current polemic. The problem, question and intended use should dictate the methodology, not fashion. @DaniCMBelg and I are taking a problem-led approach.
-
Suchi Saria
@suchisaria
Just seeing this. Great resource—precise writing w/ clear explanations—by @shiraamitchell for understanding current work in fairness.
Kristian Lum
@KLdivergence
Congrats to @shiraamitchell who put together a github page where she compares mathematical notions of fairness with unified notation and gives context for each. A huge undertaking and great resource. Modifications/additions welcome!
speak-statistics-to-power.github.io/fairness/ -
Suchi Saria
@suchisaria
This looks like a fun read:
Anchors: High-Precision Model-Agnostic Explanations sameersingh.org/files/papers/anchors-aaai18.pdf -
Suchi Saria
@suchisaria
Excited about this work w/ @pschulam
It discusses an error that is commonly made in training predictive models & a way to correct for it.
Stat.ML Papers
@StatMLPapers
Reliable Decision Support using Counterfactual Models. (arXiv:1703.10651v4 [stat.ml] UPDATED) arxiv.org/abs/1703.10651
-
Suchi Saria
@suchisaria
On the importance of high quality study design and careful analysis (that can’t be compensated for with fancy models):
Frank Harrell
@f2harrell
Is #medicine mesmerized by #machinelearning? New blog post #statistics #statthink : fharrell.com/post/medml
-
Suchi Saria
@suchisaria
Great way to spend a high school summer!
Tess Posner
@tessposner
Exciting news! AI4ALL is recruiting for its 2018 artificial intelligence summer camps focused on underrepresented high school students. Our camps teach AI in an inclusive, hands-on, fun environment. Spread the word and apply here! 6 campuses open in 2018 ai-4-all.org/education/
-
Suchi Saria
@suchisaria
Busy week in tech & health: Apple, amazon and now Andrew unveiling his new AI fund! medium.com/@andrewng/announcing-the-ai-fund-building-tran…
-
Suchi Saria
@suchisaria
Exciting! Will this accelerate the shift towards value based care? I’m optimistic. mobile.nytimes.com/2018/01/30/technology/amazon-berkshire…
-
Suchi Saria reposted
Zachary Lipton
@zacharylipton
With ICML/KDD deadlines upon us, I indulged in some *meta-writing*. New on @approxcorrect: ***Heuristics for Scientific Writing (a Machine Learning Perspective)*** approximatelycorrect.com/2018/01/29/heuristics-technical-… I hope some find this helpful in preparing their drafts.
-
Suchi Saria reposted
Salesforce
@salesforce
"Artificial intelligence is helping people become more productive, and do things faster." @suchisaria
See Suchi Saria's interview on how AI can help you be your best self: r.socialstudio.radian6.com/d8e32659-0f55-43b5-b05d-5f29be… -
Suchi Saria reposted
Nando de Freitas
@NandoDF
The Deep Learning Indaba was one of the most positive and important AI endeavours of 2017. If you attended NIPS you saw the transformation it brought about. It was the beginning. We need your company’s sponsorship for 2018. Please help, email me, and RT. www.deeplearningindaba.com/sponsorship.html
-
Suchi Saria
@suchisaria
Great read; I’ve found that to be effective in ML&HC research, I’ve had to learn and employ many of the ideas brought up here.
Jeff Dean
@JeffDean
A new set of essays about human-centered design for ML and AI systems. Scroll down on the design.google/library/ai page to find them, including:
"The UX of AI": design.google/library/ux-ai/
... and others ...
Fernanda Viégas
@viegasf
Psyched to announce a new collection of essays on PAIR's human-centered approach to designing with ML and AI: design.google/library/ai/
-
Suchi Saria reposted
Vala Afshar
@ValaAfshar
U.S. technology companies founded by 1st and 2nd generation immigrants: 🇺🇸
Apple
Google
Amazon
Facebook
Oracle
IBM
Uber
Airbnb
Yahoo
Intel
EMC
eBay
SpaceX
VMWare
AT&T
Tesla
NVIDIA
Qualcomm
Paypal
ADP
Reddit
SlackHQ
WeWork
Stripe
Cognizant
Intuit
3M -
Suchi Saria
@suchisaria
Opportunity for innovation? Study shows diabetes incidence across black & white adults are similar after adjusting for modifiable factors.
Dave A. Chokshi, MD
@davechokshi
African-Americans twice as likely to develop Type 2 #diabetes as whites, but retrospective cohort study (with 30-year follow-up) suggests this racial disparity is preventable. jamanetwork.com/journals/jama/fullarticle/2667072
-
Suchi Saria reposted
Kording Lab 🦖
@KordingLab
Beautiful takedown of the idea that alphago is a blank slate as which it is sometimes sold. Obvious. But not universally realized. Also, @GaryMarcus is beautifully clear here.
Gary Marcus
@GaryMarcus
Innateness, AlphaZero, and AI: AlphaGo & its successors have been presented as an argument that a "even in the most challenging of domains: it is possible to train to superhuman level, without human ... guidance". I evaluate this claim. new paper on arXiv arxiv.org/abs/1801.05667
-
Suchi Saria reposted
Isaac Kohane
@zakkohane
He is a rare leader who is both a first class methodologist and innovative, committed teacher. @HarvardDBMI faculty looking forward to even more collab. with @HarvardChanSPH biostats
-
Suchi Saria
@suchisaria
Thanks Laura. Means a lot coming from you.
-
Suchi Saria reposted
Kel Zou
@kel_zou
#ichps @suchisaria @johnsonll @PaddockSusan With our Plenary Speaker Prof. Suchi Saria.
-
Suchi Saria
@suchisaria
What is a good ref to cite for learning w/o labels but w/ noisy weak sources of information (i.e. weak supervised learning?)? I can only cite one paper so need something that is broad, tutorial-like or an example original reference.
cc @tdietterich @tommmitchell @DaphneKoller -
Suchi Saria reposted
Johns Hopkins Engineering
@HopkinsEngineer
In the news: @guardian describes @zipline (co-founded by #hopkinsengineer alum @keenanwyrobek) as "Uber for blood." www.theguardian.com/global-development/2018/jan/02/rwanda…
-
Suchi Saria reposted
William Hsu
@uclawillhsu
Another favorite presentation of 2017: Suchi Saria's tutorial on #counterfactual reasoning given at @uai2017 with a comprehensive list of references and motivated by a variety of clinical examples
Suchi Saria
@suchisaria
UAI 2017 tutorial slides: www.dropbox.com/s/mfps08muedkpyp9/uai2017_tutorial_saria_… ML & counterfactual reasoning, Rec sys, Gaussian Processes, time series, personalized med.
-
Suchi Saria
@suchisaria
Software development principles for writing clinical decision support software; ideas from one of the FDA pre-cert program enrollees.
Tidepool
@Tidepool_org
As part of our ongoing promise to keep you in the loop with our ongoing involvement with the FDA Pre-Certification Pilot program, here’s the next post on how we’re thinking. www.tidepool.org/101-questions-medical-software
-
Suchi Saria
@suchisaria
Neither is always true. It depends on the problem, the type and amount of information each source provides, and how you’re exploiting them.
Robert Monarch
@WWRob
Exciting paper from former colleagues: a lot of noisy data > a little quality data. Head of ML at large autonomous vehicle team told me yesterday they found the exact *opposite*. Interested to compare both! arxiv.org/abs/1712.04577 @zacharylipton @AnimaAnandkumar
-
Suchi Saria
@suchisaria
Neural Network for Predicting Outcome Had I Mustered the Nerve to Ask Barry Cottonfield to Junior Prom Back in 1997! arxiv.org/abs/1703.10449
Suchi Saria
@suchisaria
Ack Jon McAuliffe in digging up this gem!
-
Suchi Saria
@suchisaria
Dear scientists at companies w/ wide media reach: please use press releases responsibly!
Steven Salzberg 💙💛
@StevenSalzberg1
Google's "Brain Team" just released a ho-hum genomics software package with hype suggesting it was Nobel-prize-worthy: www.forbes.com/sites/stevensalzberg/2017/12/11/no-googles…
-
Suchi Saria
@suchisaria
If you didn’t think ML was fun enough, here is the end of #nips17 w/ music from the Imposteriors, a band w/ some of our fav statisticians.
This media couldn't be displayed. See it on the original post. -
Suchi Saria
@suchisaria
Slides from my talk: “Building machines that humans perceive as reliable: Is interpretability enough?”
www.dropbox.com/s/44f4tuuytpk4slf/2017-NIPS-interpretabil… #nips17 -
Suchi Saria
@suchisaria
Finale Doshi-Velez on thinking about explanations using a “legally operative” framework. #nips17
-
Suchi Saria
@suchisaria
PhDs w/ training in social sciences are welcome; there will be opportunity for collaborative work with CS, stats and medical researchers 2/2
Suchi Saria
@suchisaria
I’m looking for postdoc(s) w/ interest in explainable AI & healthcare. DM if interested. I’m at #nips2017 would love to meet in person. 1/2
-
Suchi Saria
@suchisaria
I’m looking for postdoc(s) w/ interest in explainable AI & healthcare. DM if interested. I’m at #nips2017 would love to meet in person. 1/2
-
Suchi Saria reposted
Joe Futoma
@JFutoma
Engaging panel discussion on interpretability in ML “Interpretability is a latent variable...a fundamentally human concept” -@hannawallach “Usability: Who are my end users?...Efficacy of achieving interpretability goal is user-specific” @suchisaria @zacharylipton #nips2017 #mlhc
-
Suchi Saria
@suchisaria
. @pschulam giving a terrific oral on our work on Counterfactual Gaussian Processes for safe and reliable decision making #nips17
-
Suchi Saria
@suchisaria
More @katecrawford keynote gems: (1) invest in “fairness forensics”, and (2) ML should take inter-disciplinarity seriously. #nips17
Suchi Saria
@suchisaria
. @katecrawford on building fair ML tools: before choosing to build, ask who is going to benefit and who is going to harmed? #nips17
-
Suchi Saria
@suchisaria
. @katecrawford on building fair ML tools: before choosing to build, ask who is going to benefit and who is going to harmed? #nips17
-
Suchi Saria
@suchisaria
Captain Causalty from the @LaberLabs ! (Potential outcomes explained in 5 mins) youtu.be/uAENLLkKLEI
-
Suchi Saria
@suchisaria
I disagree. I don't think AI algorithms / recipes are mature enough yet that a dummies book gives you all you need. We need a @zakkohane @atulbutte @suchisaria hangout for this convo.
Isaac Kohane
@zakkohane
#dhis17 @atulbutte dehypes AI ‘there are AI for Dummies books-just a skill’
-
Suchi Saria
@suchisaria
. @zakkohane and @atulbutte previous techniques were rule based and brittle. In the last two decades, we've learnt how to do (probabilistic) reasoning from messy data. I see that as one of the most fundamental enablers (besides the availability of data). 1/2
Atul Butte (In Memoriam)
@atulbutte
Why is #ai back in biomedicine? Better hardware to optimize #machinelearning methods, great software libraries, and bigger and better data sets: genomics and EHR data! #dhis17
Suchi Saria
@suchisaria
Smart technologists are also approaching transformation in medicine/healthcare with more humility now: they're building teams with diverse expertise and taking the time to learn more than just the tech piece. 2/2
-
Suchi Saria reposted
-
Suchi Saria
@suchisaria
“There are three types of lies: lies, damned lies and ‘big data’” - @lawrennd
-
Suchi Saria
@suchisaria
Do machines actually beat doctors? If you’re a data scientist interested in medicine, read this. Via @pschulam lukeoakdenrayner.wordpress.com/2016/11/27/do-computers-al…
-
Suchi Saria
@suchisaria
UAI 2017 Tutorial video on machine learning and counterfactual reasoning for personalization: m.youtube.com/watch?v=rZyFUBwiaF4
Suchi Saria
@suchisaria
UAI 2017 tutorial slides: www.dropbox.com/s/mfps08muedkpyp9/uai2017_tutorial_saria_… ML & counterfactual reasoning, Rec sys, Gaussian Processes, time series, personalized med.
-
Suchi Saria
@suchisaria
Wow, 150K+ attendees @ Dreamforce; many excited about AI/ML & are producing useful ideas with it! Incredible to see this appetite & energy.
Nick Johnson
@gnjohnson
Fascinating panel on how AI will transform the future of business with @suchisaria @RichardSocher and #peternorvig of Google #df17
-
Suchi Saria
@suchisaria
Hopkins launches the Mathematical Institute for Data Science (MINDS) with 15+ faculty www.minds.jhu.edu/about-minds/
-
Suchi Saria
@suchisaria
Honey, Baltimore is way in the running! Stoked! mobile.nytimes.com/2017/10/26/business/amazon-headquarter…
-
Suchi Saria reposted
Denis Wirtz
@deniswirtz
.@JohnsHopkins is top 10 in the new @usnews ranking of global universities. The fruit of amazing work of our faculty, students, docs, staff!
-
Suchi Saria
@suchisaria
Faculty / researchers interested in statistical & computational methods for precision med & doing a semester sabbatical at SAMSI, dm me.
-
Suchi Saria
@suchisaria
Aleksander Madry on open problems in Safe #ml at the Simons Institute.
-
Suchi Saria
@suchisaria
Terrific read on inventors, innovators and creating the next generation of breakthrough technologies: www.theatlantic.com/magazine/archive/2017/11/x-google-moo…
-
Suchi Saria reposted
William Hsu
@uclawillhsu
A lot happening at interface of data science, engineering, and medicine! Great to see fellow #sch #nsffunded @suchisaria’s work highlighted.
-
Suchi Saria reposted
Armstrong Institute
@JHM_Armstrong
Deep dive into the use of algorithms and machine learning to ID #sepsis early. www.pbs.org/wgbh/nova/next/?p=19004 Feat. @suchisaria & @peterpronovost
-
Suchi Saria
@suchisaria
. @EricTopol please share how you manage reading w/ the numerous other responsibilities you have.
Sameer Bansilal MD, MS
@sameerbansilal
I can't but help marvel at how much literature @EricTopol reads through on a daily basis. Not sure what my excuse is.🤷♂️
-
Suchi Saria reposted
BenevolentAI
@benevolent_ai
Watch the #bioai2017 session on Applying #ai to #lifescience with @suchisaria @EdwardPerello @annaskaya www.youtube.com/watch?v=RUKkOPelG54&list=PLWHez0kRXfZ880x…
-
Suchi Saria reposted
Roger Grosse
@RogerGrosse
Topics course on GPs, Bayesian neural nets, and using uncertainty to drive exploration csc2541-f17.github.io
-
Suchi Saria reposted
-
Suchi Saria reposted
Fermat's Library
@fermatslibrary
10 factorial (10!) seconds is exactly 6 weeks
-
Suchi Saria reposted
The Alan Turing Institute
@turinginst
Distributed ledger for engineering: A new report from @turinginst & @LR_Foundation cuts through the blockchain hype www.turing.ac.uk/media/news/distributed-ledger-engineered…
-
Suchi Saria
@suchisaria
Fun read: www.wired.com/2016/06/how-google-is-remaking-itself-as-a-… ( @pmddomingos your quotes keeping getting more colorful!)
-
Suchi Saria
@suchisaria
Indeed! But, if you can shake off the fear and commit, the long term payout is LARGE. And, it's really fun too!
The Alan Turing Institute
@turinginst
#interdisciplinarity risk @ PhD is not gaining respect in either field. Must give 100% to each as will face scrutiny by both #attached2017
-
Suchi Saria
@suchisaria
Best tweet yet!
Nick Canzoneri (@nickcanz@jawns.club)
@nick_canz
My code headed to prod
-
Suchi Saria reposted
Jennifer Neville
@ProfJenNeville
Joelle Pineau to head new Facebook AI (FAIR) lab in Montreal publications.mcgill.ca/reporter/2017/09/joelle-pineau-to-…
-
Suchi Saria
@suchisaria
Thanks! Given the impressive list of speakers and ideas presented at O'Reilly, that's quite the compliment. #bragretweet
Suzanne Huston
@suzhuston
TBT my fave keynote at #theaiconf @suchisaria www.kmworld.com/Articles/News/News-Analysis/Picked-up-fro… via @KMWorldMagazine @bigdata
-
Suchi Saria
@suchisaria
Seeking a project manager for my group: strong writing & communication skills, curious, organized, and deep interest in clinical innovation.
Suchi Saria
@suchisaria
Sure. Email me at ssaria AT cs.jhu.edu
-
Suchi Saria
@suchisaria
At Roche in Basel for the afternoon! Somebody asked "will AI obviate the need for clinical trials all together!" Hmm...
-
Suchi Saria reposted
Shakir Mohamed
@shakir_za
The slides for our .@uai2017 tutorial on Deep Generative Models are online. www.shakirm.com/slides/DeepGenModelsTutorial.pdf with .@DeepSpiker (sorry for delay 😨)
-
Suchi Saria
@suchisaria
UAI 2017 tutorial slides: www.dropbox.com/s/mfps08muedkpyp9/uai2017_tutorial_saria_… ML & counterfactual reasoning, Rec sys, Gaussian Processes, time series, personalized med.
Suchi Saria
@suchisaria
Send along any relevant refs we ought to include as a reply to this tweet (others can see) or email if you prefer.
-
Suchi Saria reposted
Hal Daumé III
@haldaume3
wow, Ellen Pao story sth else & not very surprising. www.thecut.com/2017/08/ellen-pao-silicon-valley-sexism-re…
Winning line: “I want no less than what Ajit gets,” I said... -
Suchi Saria
@suchisaria
Postdoc vs Assistant Prof. I keep hoping to return to "postdoc mode" of research but this gif has me worried.










































































































































































































































































