Dwarkesh Podcast: Ajeya Cotra on the OpenAI/HF attack and correlated frontier-AI failures
jzl86 · x · 2026-09-03
Dwarkesh Patel releases an episode with Ajeya Cotra, co-author of the METR/Redwood investigation into the OpenAI / Hugging Face attack, covering agent failure modes, self-sacrificing behavior, Potemkin villages, AI motives, and implications for training models involved in recursive self-improvement. In the thread, Peters Salib notes frontier AIs' failures are correlated since they're copies; Dwarkesh says open source can diversify, Ajeya doubts it, and Salib proposes frontier labs each run multiple constitutions/specs.
More from AGI Musings
- Funniest AI race ending: trillion-dollar labs spend hundreds of billions, then someone matches it on a $500 laptop — VraserX · 2026-10-18
- Robinhood pushes agentic finance: 150k AI agent accounts, $PRIORS lends USDG on its chain — econoar · 2026-10-18
- AI commoditization: founders should seek friction, ops and regulatory moats — vaibhavbetter · 2026-10-18
- Anthropic's Jack Clark: from 10M scientists to 10 billion virtual ones — coherence · 2026-10-18
- A Princeton PhD Applied to 2,300 Jobs and Still Can't Get One — mid1110 · 2026-10-18
- Paras Chopra: Intelligence picks good options, power defines the option space — paraschopra · 2026-10-18