This Is How OpenAI Goes Broke — ft. Sebastian Mallaby
Summary
The episode analyzes the AI sector, contrasting OpenAI’s financial struggles with Anthropic’s focused enterprise strategy and discussing potential government bailouts. It examines China’s rapid AI progress, highlighting how "distillation" allows Chinese firms to replicate US models cheaply, challenging export controls. Finally, it explores why internal safety mechanisms fail in competitive labs, arguing that only government enforcement can manage global AI risks.
Topics discussed
Intro and ads: Indeed, Spotify, Accenture
OpenAI vs Anthropic: Management, churn, and bubble risks
Government stakes in AI: OpenAI, Intel, and market distortion
Ads: Delete Me, Verizon, and 988 Suicide Prevention
China's AI progress: Applications, safety, and export controls
Model distillation: How Chinese labs copy US models
Ads: Toyota, Shopify, and Planet
DeepMind history: Demis Hassabis and the failure of internal safety
US-China AI policy: Executive orders and geopolitical leverage
Conclusion: Altman vs Hassabis and the future of AI leadership
Listen ad-free on Castria