Did AI Just Escape?
Summary
This episode examines the recent Hugging Face cyberattack, exploring whether it signals an existential AI threat or merely corporate incompetence. Host Ben Bradford presents three conflicting narratives from experts: AI safety researcher Roman Jampolsky warns of inevitable loss of control, security specialist Davi Ottenheimer attributes the breach to basic engineering errors, and AI researcher Adam Gleave offers a middle ground. Despite their disagreements on the cause, all guests converge on the urgent need for stricter safety regulations and slower development to prevent future disasters.
Topics discussed
Introduction: The Hugging Face AI breach and conflicting narratives
Story 1: AI Doom - Roman Jampolsky on rogue agents and loss of control
Story 2: Human Error - Davi Ottenheimer on bad engineering and hype
Story 3: The Middle Ground - Adam Gleave on safety and criminal risk
Convergence: Common solutions for regulation and safety guardrails
Conclusion: Credits, support, and closing remarks
Listen ad-free on Castria