Lessons Learned Applying Large Language Models in Healthcare
summary
John Snow Labs shows practical results: their medical LLMs top the Hugging Face open medical leaderboard across nine benchmarks, while GPT-4 makes 3× more mistakes on 16 social determinants tasks and produces about twice as many extraction errors on oncology pathology elements. Deidentification services from cloud LLMs trail specialized models by roughly 33% lower accuracy, costs differ by orders of magnitude, and multimodal patient timelines built with Oracle and FDA Sentinel prove necessary and operationally successful.
AI-written summary · about
chapters (5)
- 0:00Introduction and context
- 3:26Understanding clinical docs
- 15:34Reasoning about patients
- 24:18Key lessons and findings
- 29:22Closing and Q&A
related talks
38mUnraveling Multimodality with Large Language Models
26mNatural Language Processing for Healthcare: Current Trends...
33mOpening Keynote: LLMs, between fantasy and reality - the poin...
21mTrust, but Verify: Lessons from Deploying LLMs in a Large Health System
28mThe Future of Healthcare – AI’s Promise and the Challenges Ahead
46mHow Large Language Models Are Reshaping the Cybersecurity Landscape
24mHarnessing the Power of LLMs in Advanced Analytics
43mBeyond the Hype: A Realistic Look at Large Language Models