When AI's builders sound the alarm

Hoffman argues AI safety debates should center on bio/cyber/jobs risk and steering, not pausing or extinction-risk percentages.

• Hubinger's 10% estimate: Anthropic alignment researcher Evan Hubinger put his personal estimate of AI-driven extinction risk within a decade above 10%, not an official company forecast.
• Amodei's pacing proposal: Dario Amodei proposed embedded outside evaluators, inter-lab coordination, and international agreements to slow frontier capability growth; both Anthropic and Sam Altman's OpenAI have committed to evaluators.

Themes, sentiment and reasons are AI-written paraphrases, checked so they never quote an episode. They may contain errors. Not investment advice. Episodes remain the property of their creators.