Skip to content
arXiv· arXiv:2008.04071· Roman V. Yampolskiy·Indexed by this site · Originally published

On Controllability of AI

On Controllability of AI

AI summary

The arXiv paper "On Controllability of AI" argues that advanced AI cannot be fully controlled. The authors present arguments and supporting evidence from multiple fields, and note that whether AGI and its more advanced superintelligence can be controlled has not been formally established. The article also discusses the implications of this uncontrollability for humanity's future, AI research, and AI safety.

Read originalarxiv.org