Into the Impossible With Brian Keating

This AI Broke Every Benchmark — Then It Did Something Worse. Vivienne Ming - #551

70 snips
Apr 16, 2026
Dr. Vivienne Ming, neuroscientist, AI researcher, and author of Robot Proof, built an AI that refuses to answer and boosted human thinking. She discusses training a Socratic model that creates high-performing human-AI hybrids. Topics include why asking beats answering, how tools like GPS and GPT reshape memory, the Sexy Face project that reunited refugees, and what ‘robot-proof’ learning looks like.
Ask episode
AI Snips
Chapters
Books
Transcript
Episode notes

Use AI To Augment Then Outperform It

  • Use AI to augment thinking, not substitute it: consult the model, then ask 'how can I beat this' to remain actively engaged.
  • Vivienne Ming recommends pulling directions into your pocket and thinking how to improve them so you stay cognitively involved.

Information Exploration Paradox Reduces Breakthroughs

  • Ming identifies an information exploration paradox: more free answers reduce human exploration and risk-taking.
  • She observed scientists and teams herd around safe answers as accessible information increases, lowering breakthrough rates.

Keep A Failure Resume That Connects To Success

  • Keep a failure resume that links failures to later successes to train your brain to value error signals.
  • Ming explains ACC error signals drive learning, so record failures and how they produced insight later.
Get the Snipd Podcast app to discover more snips from this episode
Get the app