
Peter McCormack interviews Chris Summerfield about the implications of artificial intelligence on human cognition and the challenges it presents.
I talk to Chris Summerfield about what artificial intelligence reveals about the human mind, and what the brain still does better than today’s AI systems. We discuss why LLMs feel human, how models are trained to be helpful and agreeable, why memory and continual learning remain major unsolved problems in AI, and how sleep helps humans consolidate knowledge. They also explore agentic AI systems, rogue behaviour, reward hacking, misalignment, superintelligence, education, jobs, and whether AI will make humans more capable or less in control. - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - TIMESTAMPS: 00:00:00 - Can AI Learn Like Biology? 00:01:29 - Is AI Becoming Human? 00:04:31 - What AI Reveals About the Brain 00:06:01 - Why LLMs Don’t Remember Like Humans 00:09:38 - Sleep and Human Memory 00:13:46 - The Biggest Unsolved Problem in AI 00:15:34 - From DeepMind to Modern AI 00:20:49 - Are Humans Just Predicting the Next Word? 00:23:25 - Do We Understand What LLMs Are Doing? 00:24:16 - One of My AIs Went Rogue 00:28:07 - Misalignment and Reward Hacking 00:33:18 - Could AI Develop Self-Interest? 00:36:24 - What Happens When AIs…
Host: Peter McCormack
Guest: Chris Summerfield
Organizations: DeepMind
Explore listener stats, chart rankings, contacts and more on the The Peter McCormack Show podcast page.