Summary

Zhu speaks about the nature of ontological crisis – a state of reality shift for either human or machine. He then transitions into the ontological anti-crisis, and how to use such a phenomenon to increase safety for artificial general intelligence.

Presenters

Presentation: Ontological anti-crisis and AI safety

Transcript
  • The center for safe artificial general intelligence uses partnerships in China to investigate methodologies for developing AI safely.
  • The (extreme) change of model of the reality for an entity

  • Humans, machines, and posthumans can undergo ontological crisis.

  • Decision diagram – machine ontological crisis is likely easier to fix.

  • The ontological hierarchy is a product of how ontological progress is built into the human or machine condition.

  • Ontological completeness and incompleteness Theses for safe AGI

  • Ontological anti-crisis is a systematic way to design an ontological structure which takes change into account.  It can help humans and machines understand each other.

  • OAC may be able to help with AI safety

  • Iterations of the OAC theory may be used to improve completeness and create safer interactions with AI