AI-written NIH grant abstracts pass the vibe check, but humans still spot the real one
TuhinChakr · x · 2026-11-12
Andrew White's team released a demo comparing model-written NIH grant abstracts with real funded proposals. Joel Chan found the AI versions less LLM-y yet still predicted the real one 6/6 — his vibe: LLM writing remains suspiciously 'clean'.
More from Fun
- Augmenting solo Descent 2e tabletop sessions with AI, CV, IR and a document camera — TheZachMueller · 2026-11-12
- A look at Qwen's body representation inside an LLM — Sauers_ · 2026-11-12
- "This is what an LLM's body representation looks like" — Sauers_ · 2026-11-12
- AIE Code Summit's "Closing Keynote" Goes to the Slot When Everyone's Already Flying Out — trq212 · 2026-11-12
- An interactive visualization that makes Bayes' theorem something you can actually see — burny_tech · 2026-11-12
- DeliveryKit Lands Its First $29 Subscription Payment via Stripe — ow · 2026-11-12