Fewer than 5% of FDA-cleared AI/ML devices are proven via randomized trials
jonc101x · x · 2026-10-01
Physician Jon Carter notes that fewer than 5% of FDA-cleared AI/ML medical devices are validated through randomized trials the way new drugs are — while pointing out that medical textbooks and review articles are routinely published and used without that level of scrutiny either.
His thread argues credible AI progress is happening in clinical tasks with clear boundaries and measurable endpoints, but broad clinical autonomy and "AI doctors" aren't ready for primetime. Full review published in BMJ Digital Health & AI.
More from AGI Musings
- Cambridge's SAM-Brain3D surpasses existing methods for early brain disease detection — lawrennd · 2026-10-01
- The Real Bottleneck in AI Drug Design: In Vivo Behavior Still Needs Animal Testing, AI Can't Predict It — anshulkundaje · 2026-10-01
- geohot essay asks: when aligned superintelligence knows you better than you, what is freedom? — burny_tech · 2026-10-01
- Designer essay: AI gets me to 99% ten times faster — and the 1% I keep accepting is adding up — round · 2026-10-01
- Anil Seth's Prize-Winning Essay Argues Consciousness Belongs to Life, Not Computation — jmugan · 2026-10-01
- Four years, no progress: Redditors debate why LLMs still can't do nuance — Hatrct · 2026-10-01