SITUATION DETECTED: Anthropic co-founder Jack Clark says he now believes it is likely that full recursive self-improvement will become possible in 2027.

the researchers at all the labs would tell you in private and sometimes in public that alignment has not been solved. if this is true, then racing into RSI is a folly that is unprecedented in world history. we will need to significantly slow the frontier of research very soon

If our doubling trend continued and the “actual” taste involved in frontier AI research had the same doubling time as the one that TasteVal measured, AI takeoff expectations would be faster: the AI Futures Model's probability of a taste-only singularity would rise from 51% to 88%, and its median date for superintelligence moves about 18 months earlier, from mid-2030 to late 2028. But the "actual" taste involved in frontier AI research may double at a different rate, since (1) our tasks run on a single GPU with fast feedback loops while frontier experiments take weeks on big clusters, (2) we don't measure choosing which problems to work on, and (3) our trend may not continue.

Minerva author on AI solving math: - IMO gold by 2026 seems reasonable - superhuman math in 2026 not crazy - auto-formalizing is unimpressive to mathematicians as most important theorems are hard to formalize

We’re releasing a broad range of new mathematical results produced by an internal frontier model. We’ve been consulting with the independent Advisory Group on Mathematics and Artificial Intelligence at the Institute for Advanced Study, and we have drawn on their advice and public recommendations to inform how we release these results. github.com/openai/math

My testimony at the hearing on AI at @NYCCouncil