The World as You’ll Know It: The Future Of Aging Watson Part 2: How IBM’s Big Bet Failed
5 snips
May 2, 2023 AI Snips
Chapters
Transcript
Episode notes
MSK’s Early Watson Collaboration Went Off Script
- Dr. Caroline recounted MSK's early collaboration with IBM that began with high hopes but quickly diverged from research goals.
- IBM corporate pushed for data and clinician time while MSK wanted a genuine research partnership, causing mounting frustration and eventual removal of Dr. Caroline from the project.
EMRs Hide Crucial Clinical Context
- Electronic medical records prioritize billing, leaving key clinical details in free-text notes that machines struggle to interpret.
- Watson failed partly because it couldn't reliably read or know when it lacked sufficient information and still made recommendations.
Data Prep Is Most Of Medical AI
- Data cleaning and contextual understanding dominate medical AI work and were under-resourced in Watson's project.
- Machines can detect spurious patterns without context, leading to dangerous or useless conclusions in healthcare settings.