Dario Amodei Interview

Dario extrapolates capability curves: from high school to undergraduate to PhD-level AI. He sees AGI arriving by 2026-2027, though worlds where it takes 100 years are rapidly decreasing.

Discussion of how AI models are deployed in thousands or tens of thousands of instances today, and how within 2-3 years clusters will support millions of AI instances.

Dario's biggest worry isn't existential risk but economic concentration of power. AI increases the amount of power in the world, and concentrating and abusing that power can do immeasurable damage.

Philosophical exploration of whether human meaning survives in a world where AI can do everything better. Dario is optimistic about meaning but worried about power dynamics.

How Anthropic differs from OpenAI in its approach to safety, Constitutional AI, responsible scaling policies, and the RSP framework.

Extended discussion covering geopolitical implications of AI, the race between nations, and what humanity should be doing now to prepare.