Dwarkesh Podcast

Embracing Painful Engineering Bottlenecks for AI and Space Scale

Scaling AI requires confronting painful, long-term engineering challenges, from energy constraints to advanced manufacturing, revealing space as a surprising solution for compute power.

Brain's Complex Cost Functions Drive Efficient Learning Beyond AI

AI's path to true intelligence is blocked by its misunderstanding of the brain's encoded reward functions. Discover how evolution's "secret sauce" offers a blueprint for more efficient AI.

Human-Like Learning, Not RL, Drives Future AI Progress

Current AI training methods are flawed if models can learn on the job like humans. This reliance on pre-baked skills may soon be obsolete, highlighting a significant gap in true AI generalization and adaptability.

Internal Soviet Flaws Drove Collapse More Than External Pressures

Internal systemic flaws, not external pressure, collapsed the Soviet Union. Gorbachev's reforms and the Helsinki Accords inadvertently accelerated its demise, revealing its inherent unsustainability.

AI's Scaling Plateau: The Urgent Return to Foundational Research

AI is transitioning from scaling to research, facing a generalization gap between benchmarks and real-world utility, demanding novel training for true intelligence.

Microsoft's AI Infrastructure: Scaling for the Cognitive Revolution

Microsoft invests heavily in AI infrastructure, evolving business models beyond SaaS and integrating AI agents to transform computing and empower human and autonomous capabilities.

Russia's Century-Long Sabotage of China's Rise

The Decade of Agents: AGI A Decade Away, Imitation Learning Reigns

Eukaryotic Leap: The Improbable Spark of Complex Life