Another one from the Bleibach Totentanz - or Dance of Death. This scene shows death coming for the physician - and you can tell it’s a physician because of the urine flask, which used to be their primary diagnostic tool. It was believed that the colour of the urine would indicate the disease

Remember when you first learned about genetics at school? All those fascinating examples of human traits that are each apparently determined by just a single gene? Time to check in on some of your favourites to see how they’re doing. 🧬🧵🧪 1/n

2 important papers @nature.com today indicate the potential for A.I. to improve diagnostic accuracy and overall clinical performance for patient management and interactions nature.com/articles/s41...

Editorial by @adamrodmanmd.bsky.social and colleagues: When It Comes to Benchmarks, Humans Are the Only Way nejm.ai/4hMrAM9 #AI #MedSky #MLSky

IMO this is the best argument for robust benchmarks. We have no way of studying new models other than as if they were natural phenomena. Clinical trials are incredibly important -- but with this rate of change so is benchmarking for the foreseeable future.

If you're a member of the American College of Physicians and want to learn more about LLMs and their implications for medicine, we put together a webinar! link: www.acponline.org/clinical-inf...

Can a #MedEd #Podcast improve your OSCE score? Yes—and now it’s published! This RCT by V.Dupont et al. shows that listening to #NephrOdio leads to: 🎧 +3.5 OSCE score boost 💪 Better self-perception of clinical skills (42% vs 16%) ✅ 96% would use the podcast again 👉🏼 doi.org/10.1186/s129...

🚨 New paper 🚨 Excited to share new capabilities for AMIE our conversational medical AI @googIe.blog We advance AMIE beyond diagnosis 🩺 towards treatment 💊 over time w/ guidelines 📃📃📃 - Dive into 🔗 research.google/blog/from-di... with my amazing co-author @adamrodmanmd.bsky.social!

Medical education 🌶️🔥take: the threat of cognitive deskilling from genAI technologies is the #1 things medical educators need to be talking about right now.

We are excited to share our latest results on AMIE, Google Research and Google DeepMind's conversational medical AI. TL;DR – in a blinded OSCE, AMIE was equivalent or better at diagnosis – but also at making management decisions. This is a big deal! 🧵⬇️

AI could be used to prescribe medications to patients -- if a new bill makes its way through Congress. 🤖 💊 What do doctors think about the proposed legislation? The technology isn't there yet, @adamrodmanmd.bsky.social told MedPage Today. www.medpagetoday.com/special-repo...

How do you know a tree when you see it? 🧵1/8

Quick Saturday AM reminder to contact your congresspeople if you're upset about the sudden cuts in NIH indirects (or, like, anything else).

There is a lot of buzz about our new paper in Nature Medicine on the effects of LLMs (GPT-4) on physician management reasoning! I had TONS of fun working on this -- but what it MEANS requires some unpacking. A 🧵⬇️ bsky.app/profile/ucsf...

A new study in @natmedicine.bsky.social finds GPT-4 can enhance physicians’ clinical decision-making, improving open-ended management reasoning. While promising, experts stress the need for human oversight to mitigate risks like AI bias and over-reliance. Read more: www.nature.com/articles/s41...

Any studies on the relative sizes of lower extremities in heart failure exacerbation? @sargsyanz.bsky.social (And if no, anyone want to do one?) Best research ideas come from things that annoy you after all!

Cover reveal thanks to the clever people @yalebooks.bsky.social

I don't have anything snide or snarky to say here. I just think it's worth reading this article. Gift link.

Some clinical documentation is busywork and some is the byproduct of thinking critically about the patient. There are an awful lot of people trying to use AI to do away with the latter.

I think often about this experiment at Harvard showing active learning upped test scores by 33%... but students thought they were learning more from non-active lectures The problem: being challenged is the key to learning BUT we hate challenges, they are hard & make you realize how little you know.

also:

It’s happening tomorrow !! #iMED2025 Register today! bit.ly/3P7HUvb

🚀The return of BERT, LLMs that beat physicians?!, and the dawn of agents to simulate users. Presenting Santa's nice list for AI! 🎅 - ModernBERT - Superhuman performance of LLMs against Physicians (discussion with author @adamrodmanmd.bsky.social ) - LMagent: A Large-scale Multimodal Agents Society

Independent evaluations of OpenAI’s o3 suggest that it passed math & reasoning benchmarks that were previously considered far out of reach for AI including achieving a score on ARC-AGI that was associated with actually achieving AGI (though the creators of the benchmark don’t think it o3 is AGI)

We have two vacancies for a postdoctoral and PhD researcher at Maastricht University. They will work on the development and application of value-of-information methods for the validation of clinical risk prediction models and medical AI. links below

Our preprint out today! I just don't have the stamina to make threads on both platforms 😅

I really wish there was more interest in benchmarking AI models on real tasks. The fact that there are not 30 different benchmarks from different organizations in medicine, in law, in advice quality, etc. is a big shame. People are using systems for these things anyway & we don’t know implications.

I'm crabby today but I wish we could have a little less pearl-clutching over "the children are citing AI hallucinations" and a little more over "tenured academics are citing documents that cite documents that cite documents that cite documents that don't support their claims."

I think a fun teaching exercise for medical students would be to each take a common clinical pearl and trace the citations until you find the primary data it's based on.

Our HMS Digital Education and AI conference is back! We're having a prompt competition this year -- how do you use AI to enhance your patient care, home life, or what you do for fun? So sign up! Competition: forms.gle/YwEHkwk884sZ... Registration: cmecatalog.hms.harvard.edu/digital-educ...

Tele-ICU consultation might for these reasons prove to be a fertile ground for investigating precisely the sort of multi-agent interaction problems that plague human-AI HCI. How do you effectively incorporate info and decisions from any outside source, human or otherwise? @adamrodmanmd.bsky.social

This is a horrifying story.

Anyone with any recommended readings about the connections between Ayurveda and humoral medicine pre-Alexander? (there may be no scholarship...)

THREAD: Research we should be thankful for. The chances that we and everyone we know and love would be here today, alive, healthy and thriving without the advances of science and technology are basically zero. Today we look through the research and scientists to whom we owe everything. 1/19

I was on CNN last night, talking about our latest study! www.cnn.com/2024/11/27/h...

Where does this assumption that doctors should just passively be the "human in the loop" for AI systems come from? Either AI systems outperform us at a task and we use them instead; or we work better together and we do that; or humans are better and we do that.

If you somehow found me on Bluesky and you're wondering why I haven't released a new episode of Bedside Rounds in over a year, I wrote an article about it in Harvard Magazine! link: magazine.hms.harvard.edu/articles/can...

Okay -- I'm going to try Bluesky. Not giving up Twitter yet (for better or worse a lot of the ML and AI in medicine community is there, and it's still a good way to share and discuss literature), but Twitter hasn't been good for science discussion for quite some time.