New in JMIR MedEdu: How Does ChatGPT Perform on the United States Medical Licensing Examination? The Implications of Large Language Models for medical education #mededu and Knowledge Assessment https://t.co/JoLWaTqqxD
A key goal of @gpt_index is to enable end users to ask an LLM *any* questions over their own data. Building this universal query interface is hard!
@ezhu worked on a superb evaluation 💡 of @gpt_index data structure query capabilities. Check out thread 🧵 + Colab nb below!
Our lab recently published a paper on ChatGPT's perf on Medical Licensing Exams: namely USMLE and AMBOSS (practice Qs used by med students) achieving 64.4% in Step 1 (4% over passing threshold) and capturing logical reasoning and internal info well! https://t.co/JNqCi3oYEP
MIMIC-IV was published this week. The core dataset has been out for a while, but we've just published the deidentified free-text clinical notes: 300,000+ discharge summaries and 2.5 million radiology reports!
https://t.co/WVO2PFKij1
The first day of our @YaleBiotechClub Datathon was a success! 7 teams worked on novel ways to draw insights from mutliomic datasets, @ZvxyWu (@DeepMind) gave a keynote speech on applications of ML to structural bio, and our wonderful partners and mentors from @CodeOceanHQ,
Had an awesome time at the @YaleBiotechClub Datathon sponsored by @boehringerus and @CodeOceanHQ!! Thanks to @ZvxyWu for an amazing Keynote! The teams have worked crazy hard on a multi-omics integration problem. Stay tuned for results viz/presentations! @YaleCBB
Had a great time at the @UofSCbdhsc Health Data Case Competition. Made it through a two rounds and got 3rd place! The sleepless nights were worth it, congrats to @MarkovYaro and @danpak94!!
Our most recent talk comes to us from @NicholasAder as he discusses how the forms of cells can teach us about their function and the different types of microscopy which help us view these cells! https://t.co/nXhPiEWsgn
Some clinicians work on different EHR products on a routine basis. Basic standards for EHR data elements like critical lab values can reduce the cognitive burden on clinicians & improve safety. The automotive industry can agree, & my DVD player & Keurig can agree. Why not EHRs?
Those who say #AI will replace doctors completely don't understand medicine, and those who say #AI can't replace many medical tasks don't understand #AI.