Boss-level 'bug': Rogue lunar effects upon RL-trained NN for playing the Rogue-like game Nethack (wasted many hours during my PhD: A debugging journey through a ML stack. Worthwhile thread below:

Fun conversation and change to reflect about trajectories--and cycles @nejm_ai.

I recently had the privilege of presenting some words at the 2024 Graduation Commencement of the McWilliams School of Biomedical Informatics (thanks for the invite, Dean @JiajieZhang100)!

Here is the speech I delivered, if anyone is interested:

Dear parents, families, faculty members, staff, and of course

From our People-Powered Medicine group:
Can you team up with your doctor to prevent heart disease when you have access to your **own** data? Download the People Heart Study from the App Store at bit.ly/people-heart-s… to participate.
People-Powered Medicine Isaac Kohane Raheel Sayeed, MD

Sadly retiring our Swiss-made JURA US Impressa E8 after producing more than 24 thousand shots of espresso over nearly 20 years!

It’s more than a mentorship-mentor team. I’ve witnessed here the emergence of an exciting and productive research partnership. Looking forward to future dividends

No #22 without #23!

A huge thank you to Raj, my amazing advisor of eight (!) years and the absolute GOAT of supporting my career 🙏

Low and middle income countries do not have to wait for high-income countries to address data science ethical issues including those affecting medicine. A useful peek into frameworks in Nigeria NEJM AI ai.nejm.org/stoken/default…

With open source tools now coming from EHR vendors to validate medical #AI algorithms, wondering what will be the value of complicated networks of assurance labs…

UCSF CoDEx Join us to learn more about the Patient AI Rights Initiative!

cc: Dave deBronkart Sneha Dave GRASP Enlightening Results 💡 Ysabel Duron Jill Holdren Hugo Campos Isaac Kohane Karandeep Singh Dana M. Lewis | #OpenAPS 🤖 Marlena GRASP Maryam Lustberg MD MPH

#BeLikeLight

Sula Ndousse-Fetter our brilliant Harvard Medical School IV student expertly presenting her project on the effect of Mtb lineages on risk of dissemination and/or extrapulmonary involvement of TB to an intent audience. ATS PI-TB Assembly #ATS2024 🙌 👏 🤓

Jie Yang Isaac Kohane NEJM AI Awesome work! I love the level of detail you go into with respect to the accessibility of datasets. 'Available on request' is often not meaningful in practice, and journals should refuse to accept such language in absence of clear protocols for dataset archiving and access.

Our recent paper 'Clinical Text Datasets for Medical Artificial Intelligence and Large Language Models — A Systematic Review' has been published in NEJM AI ! The lack of clinical text data is a long-term challenge/pain for #AIinMedicine and clinical LLM researchers. After

Happy to have contributed to this massive effort! Congrats James Diao and Arjun (Raj) Manrai. Definitions and measurements matter in medicine, and PFTs equations are no exception 1/2

This was my dream team for a Health and AI panel and it happened! Peter Lee Julie Yoo Isaac Kohane #fortunebrainstormhealth

RAG improves a GPT copilot in oncology guidelines questions. No RAG: 57% correct. RAG: 84% correct. Is that enough? The need for a human pilot remains, but attaching reliability evidence like this paper to a copilot can help clinicians make entrustment decisions. NEJM AI

Graduation commencement today at UC San Francisco!

Special Article: Implications of Race Adjustment in Lung-Function Equations nej.md/44K5TaC

Editorial: Beyond Diagnostics — Removing Race from Lung-Function Test Interpretation nej.md/3yssagU

#ATS2024 American Thoracic Society (ATS)

Peter Szolovits, PhD, discusses the capabilities and incomprehensibility of large language models (LLMs) and their use in medicine. Read the Perspective “Large Language Models Seem Miraculous, but Science Abhors Miracles”: nejm.ai/4aqBIqe

Fantastic work by James and prof Arjun (Raj) Manrai