Dr. John Vervaeke

What AI Can Never Be | John Vervaeke

155 snips
May 30, 2025
Can AI ever achieve true wisdom? The discussion dives into the profound differences between intelligence and wisdom, revealing the limitations of language models. It examines the essence of human cognition, emphasizing embodiment, caring, and participatory knowing. Concepts like relevance realization and the four kinds of knowing highlight why AI struggles with consciousness and understanding. The talk ultimately stresses that for machines to become wise, humanity must first cultivate its own wisdom.
Ask episode
AI Snips
Chapters
Transcript
Episode notes
INSIGHT

LLMs Solve Many Problems Narrowly

  • Large Language Models (LLMs) like ChatGPT can solve many domain problems but only within propositional knowledge.
  • LLMs break the AI silo problem but don't provide a scientific explanation of intelligence applicable to other creatures.
INSIGHT

Two Meta Problems of Intelligence

  • Intelligence involves solving two meta problems: anticipation and relevance realization.
  • Anticipation is predicting future states; relevance realization means filtering overwhelming information to what's important.
INSIGHT

Human Intelligence Predicts Across Domains

  • Human intelligence predicts performance across diverse domains like art, math, and sports.
  • This general ability contrasts with narrow AI capabilities.
Get the Snipd Podcast app to discover more snips from this episode
Get the app