✳flâneur — a map of the web's best reading

AI in 2025: gestalt — LessWrong

lesswrong.com · 11,502 words · saved by 11 readers

This is the editorial for this year’s "Shallow Review of AI Safety". (It got long enough to stand alone.) Epistemic status: subjective impressions plus one new graph plus 300 links. Huge thanks to Jaeho Lee, Jaime Sevilla, and Lexin Zhou for running lots of tests pro bono and so greatly improving the main analysis. Better, but how much? Fraser, riffing off Pueyo We now have measures which are a bit more like AGI metrics than dumb single-task static benchmarks are. What do they say? So: is the rate of change in 2025 (shaded) holding up compared to past jumps?: Ignoring the (nonrobust)[5] ECI GPT-2 rate, we can say yes: 2025 is fast, as fast as ever or more. Even though these are the best we have, we can’t defer to these numbers.[6] What else is there? One way of reconciling this mixed evidence is if things are going narrow, going dark, or going over our head. That is, if the real capabilities race narrowed to automated AI R&D specifically, most users and evaluators wouldn’t notice (espe

x AI in 2025: gestalt — LessWrong Distillation & Pedagogy Fact posts AI Frontpage 2025 Top Fifty: 11 % 248 AI in 2025: gestalt by technicalities 7th Dec 2025 24 min read 44 248 This is the editorial for this year’s " Shallow Review of AI Safety ". (It got long enough to stand alone.) Epistemic status: subjective impressions plus one new graph plus 300 links. Huge thanks to Jaeho Lee, Jaime Sevilla, and Lexin Zhou for running lots of tests pro bono and so greatly improving the main analysis . tl;dr Informed people disagree about the prospects for LLM AGI – or even just what exactly was achieved

Explore this link on the map →

saved by

related reading