GZERO World with Ian Bremmer The human cost of AI, with Geoffrey Hinton
81 snips
Dec 6, 2025 
Geoffrey Hinton, a renowned computer scientist and Nobel Prize winner, discusses the alarming pace of AI development and its societal implications. Once an AI evangelist, he now warns of job loss, inequality, and the potential existential threat posed by superintelligent systems. Hinton delves into the complexities of large language models, safety risks, and the need for AI to be designed with human 'maternal instincts.' He proposes training AIs to model good behavior to mitigate catastrophic outcomes while highlighting the urgency of addressing these issues.
AI Snips
Chapters
Transcript
Episode notes
AI Could Surpass Humans Within Two Decades
- Geoffrey Hinton thinks AI will likely surpass human intelligence within ~20 years.
- He warns a significant chance exists that smarter AIs could take over if unchecked.
We Program Learning, Not Final Behavior
- Hinton explains we program learning rules, not the final behavior of large language models.
- The trillion weights learned from data determine why a model outputs a specific answer.
Factor Social Disruption Into AI Economics
- Policymakers must factor in huge social disruption from productivity gains concentrated among owners.
- Without sharing wealth, mass automation will cause severe inequality and instability.