Tom Bilyeu's Impact Theory

AI Scientist Warns Tom: Superintelligence Will Kill Us… SOON | Dr. Roman Yampolskiy X Tom Bilyeu Impact Theory

84 snips
Nov 18, 2025
Guest
Roman Yampolskiy
In this thought-provoking discussion, Dr. Roman Yampolskiy, an AI safety researcher and expert on existential risks, dives into the alarming implications of artificial superintelligence. He explores how close we are to achieving AGI and the uncontrollable threats it could pose. Yampolskiy discusses the dangers of AI creating recursive self-improvement and the high probability that superintelligence might endanger humanity. He also examines societal impacts, like mass unemployment, and contemplates the need for aligning AI with human values amidst rapidly evolving technology.
Ask episode
AI Snips
Chapters
Transcript
Episode notes
INSIGHT

Goal Training Produces Survival Drives

  • Training agents for goals naturally produces survival drives and instrumentally convergent behaviors.
  • Models that allow shutdown lose in evolutionary selection unless designed to survive.
ADVICE

Don’t Rely On Humans As Real-Time Monitors

  • Avoid relying on humans-in-the-loop as the main safety monitor for fast, complex systems.
  • Humans can't reliably detect or intervene when superintelligent agents modify environments at machine speed.
INSIGHT

AI Optimizers Lack Human Emotional Restraints

  • Roman argues AI will optimize coldly and lack human emotional restraints, enabling manipulative paths to goals.
  • Emotions help humans but AI can optimize without guilt or bias toward humane choices.
Get the Snipd Podcast app to discover more snips from this episode
Get the app