"Scaling, ASI, RSI is fortune-telling, not science": Kundaje's critique of capability prognostication

anshulkundaje · x · 2026-09-09

Continuing his thread, Anshul Kundaje says he's all ears if anyone can provide a map of how new capabilities might be achieved — but just invoking words like scaling, ASI, or recursive self-improvement is "extremely dissatisfying."

Nothing wrong with doing so, he notes, but it feels like fortune-telling instead of science, echoing his earlier point that the nature of available data sets hard limits on what models can do.

Related event: Stanford professors clash over AI limits: data bottlenecks vs a recent convert's 'it just works'(10 posts)→

Original post →

More from AGI Musings

AGI Musings channel →