OpenAI Whistleblower FINALLY Speaks: “AI Has A 70% Chance Of Going Horribly Wrong!“
Summary
Steven Bartlett interviews Daniel Kokotajlo, a former OpenAI researcher who resigned over safety concerns and refused a $2 million payout to avoid signing an anti-disparagement clause. Kokotajlo warns that AI companies are racing toward superintelligence by 2027, risking catastrophic loss of control or extreme power concentration. He argues that current alignment efforts are insufficient, urging public awareness of the existential threats posed by unchecked AI development.
Topics discussed
Introduction and Helix sponsorship
AI risks, job displacement, and expert consensus
Daniel Kokotajlo's background and OpenAI experience
The OpenAI equity controversy and internal dynamics
Publishing the AI 2027 forecast and rapid progress
Defining superintelligence and neural net mechanics
Anthropic's strategy and CEO incentives
Job automation timelines and economic impact
AI 2040 Plan A: A recommended safe path
International regulation and the AI 2040 scenario
Citizen's checks and societal transition
Long-term future: Space expansion and civilization
Personal reflections, closing thoughts, and sponsor
Listen ad-free on Castria