Andrew Huberman· PhD
The GPT4 was equal or better on most of the factual questions analytically, already, and it's not even a specifically trained medical AI, but it was overwhelmingly better on empathy.
Loading…
The headline is broadly defensible, but the qualifications matter. Effect sizes vary by population, the strongest claims rest on shorter trials, and credible voices push back on how it's typically framed.
The GPT4 was equal or better on most of the factual questions analytically, already, and it's not even a specifically trained medical AI, but it was overwhelmingly better on empathy.
Every Sunday: the week’s new conflicts and verdict changes — and nothing else.
Native comments, Twitter mentions, and Reddit threads about this claim — surfaced together so the conversation isn't fragmented across platforms.
Bookmarking — the dossier-vs-overview split is the right call. Most of the time I want overview; sometimes I want receipts.
Would love a "what would change this verdict" RSS feed. Sign me up if it exists.