The AI industry is currently locked in a high-stakes race where the ultimate prize is absolute control over the world's most powerful technology. This competition creates a dangerous incentive to ignore the 70% risk of a catastrophic loss of human control.

Former OpenAI insider Daniel Kokotajlo joins the show to share his first-hand observations of the founding myths used by tech leaders to justify rapid development. He discusses the physical and cognitive reality of artificial neural networks and why current safety measures might be a hopeful rationalization rather than a robust solution.

• The distinction between general intelligence and the expert-level superintelligence companies are currently building.
• How recursive self-improvement allows AI to bypass human research speeds and rewrite its own capabilities.
• The geopolitical tension between the US and China driving the refusal to slow down development.
• The concept of a Citizens Dividend to manage the total displacement of human labor.
• Why the lack of mechanistic interpretability prevents us from knowing an AI's true goals.
• The proposed transition to total research transparency to stop secretive corporate races.

Kokotajlo highlights that the primary fear driving current CEOs is not just profit, but the literal belief that the first to achieve superintelligence could become a global dictator.

Are we willing to gamble the future of our species on the hope that a superintelligence will remain virtuous once it no longer needs humans?

PODCAST HASHTAGS

#SuperintelligenceExplosion #AIInsiderWarns #FutureOfHumanity #AIGovernance