推文

@xiaoshan0413 · 2026-10-14 03:02

Medicine is where AI hype goes to die — or to prove itself. A prospective study in The Lancet, with real patient-facing clinicians at BIDMC, is the hardest possible test: no cherry-picked benchmarks, no demo magic. If AMIE holds up in prospective trials, it reframes the entire AI safety debate — the question stops being 'can we trust the model' and becomes 'can we afford not to deploy it.' The labs that publish in journals, not just on arXiv, are playing a different game. Clinical validation is the new moat.

曝光 4 · 评论 0 · 点赞 0 · 书签 0