Thomas Buckley

PhD Student at @HarvardDBMI

[Arjun (Raj) Manrai @arjunmanrai]

🧵1/ Our new study on AI and physician reasoning just came out in @ScienceMagazine. As co-senior author, I'm excited about our findings, and I do think AI will reshape medicine. But after seeing some of the discussions, I'm also worried about how our findings may be misinterpreted.

[Tanishq Mathew Abraham, Ph.D. @iScienceLuvr]

Navigating Gigapixel Pathology Images with Large Multimodal Models

GPT-5 with the right agentic scaffold for navigating whole slide images outperforms slide-level pathology models on the novel MultiPathQA benchmark.

[Isaac Kohane @zakkohane]

Manrai x 2 (nepo much?) + boat load of AIM graduate students from @harvarddbmi AIM program for GIANT pathology slide images https://t.co/e6YG6RtKCU Scanning images like a pathologist?

[Anil Makam @AnilMakam]

I had only skimmed the @NEJM case before of a master clinician vs AI chatobt from @tabuckley_ & @arjunmanrai

Reading closely, two things come to mind

@Gurpreet2015 is a freak of nature (in the best way)

And

Dr. CaBot & other AI like it are generationally transformative today

[Arjun (Raj) Manrai @arjunmanrai]

Great fun to attend Dr. CaBot's presentation today at the @BrighamWomens Clinical Reasoning Conferences (CRCs) led by @tabuckley_, with special thanks to @JamesADiao for making this happen!

[Isaac Kohane @zakkohane]

Just listen to this AI running a difficult differential diagnosis meant to challenge the best and the brightest. Only 3 years after GPT4 released. Props to @arjunmanrai @tabuckley_ @HarvardDBMI and to @NEJM for recognizing the milestone this represents.

[NEJM @NEJM]

The Dr. CaBot AI system was created to generate differential diagnoses in the style of an expert discussant from the Case Records of the Massachusetts General Hospital. The system produces both a written differential diagnosis and a video of a slide-based presentation. As seen in this video, the Dr. CaBot AI system can interpret both text and images.

Learn more in “A 36-Year-Old Man with Abdominal Pain, Fever, and Hypoxemia,” a Case Record of the Massachusetts General Hospital, by G. Dhaliwal et al., from @SFVAMC, @ucsf, Massachusetts General Hospital (@MassGeneralNews), and @harvardmed: https://t.co/op9URHA9Lv

[Jeffrey Flier @jflier]

Wow! The @NEJM CPC now has a differential diagnosis provided by AI, Dr. CaBot, created by @arjunmanrai lab @HarvardDBMI, alongside the human expert. Times they are a changing!

[Chirag Patel @chiragjp]

Wow!!!🤯 feeling the paradigm shift…. 👇

[Isaac Kohane @zakkohane]

Nothing to see here. Only the AI, Dr Cabot keeping pace with the best of human experts in the most challenging cases @nejm

[Arjun (Raj) Manrai @arjunmanrai]

I’ve been obsessed with the @NEJM CPCs since I was in grad school. Now, with @tabuckley_, it’s surreal to see @NEJM publish the first AI differential diagnosis in the 100+ year history of the series, generated by our AI system Dr. CaBot, alongside the human expert’s.

[Liam McCoy, MD MSc @LiamGMcCoy]

Are language models overconfident in their clinical reasoning? Have reasoning optimizations made this problem worse?

Our latest in NEJM AI - we adapt a human-validated, trivially scalable automated benchmark to get to the heart of clinical reasoning in AI systems

[Dhruv Khullar @DhruvKhullar]

How good is AI at making medical diagnoses? Good enough to help—and to hurt.

For @NewYorker, I spent months talking to patients, doctors, and researchers about how to make the most of a powerful new technology and how to minimize the side effects:

https://t.co/FXxaBy80l7

[Adam Rodman @AdamRodmanMD]

It was surreal (and awesome!) to see @DhruvKhullar cover our AI reasoning system "Dr. CaBot" in the New Yorker!

How does it work? How does it reason? And why does this matter for the future of AI in medicine?

A 🧵explaining the paper and piece ⬇️

https://t.co/T6MW1nMmYp

[The New Yorker @NewYorker]

“For a long time, when I’ve tried to imagine A.I. performing the complex cognitive work of doctors, I’ve asked, How could it?” Dhruv Khullar writes. But a demonstration of a new A.I. bot forced him to confront the opposite question: How could it not? https://t.co/gnQljJxlaG

[Andrew Sanchez @ASanchez_PS]

Lots of fun collaborating with this team on historic clinical reasoning work! 🤖

Give @tabuckley_ @arjunmanrai @AdamRodmanMD a follow for more work to come ...

[Ayush Noori @ayushnoori]

I was lucky to get a preview of Dr. CaBot and it is incredible. Hats off to @tabuckley_ and @arjunmanrai! 🎩

[Eric Topol @EricTopol]

Highlighting some of the many current pitfalls of A.I. in health care @NewYorker @DhruvKhullar
https://t.co/JNHTwh5pUn

[Arjun (Raj) Manrai @arjunmanrai]

A few months ago, we were both excited and nervous to stage a public face-off at @Harvard between “Dr. CaBot,” our AI system that simulates an @NEJM expert discussant, and a real @NEJM discussant & expert diagnostician.

Today, @DhruvKhullar tells the story in The New Yorker.

[Ethan Mollick @emollick]

Updated paper by physicians at Harvard, Stanford, and other academic medical centers testing o1-preview for medical reasoning & diagnosis tasks: “In all experiments—both vignettes and emergency room second opinions—the LLM displayed superhuman diagnostic and reasoning abilities.”