Solipsistic Superintelligence is Unlikely to be Cooperative

Arxiv CS.AI (top papers)

Paper argues that reward-maximizing superintelligences will be solipsistic and uncooperative, posing coexistence risks. Authors propose a shift from capability to cooperation in AI research.

See 1 more source

The Most Terrifying Superintelligence Might Not Want to Rule Us at All.14 days agoShanbhag01Discuss

Arxiv CS.AI (top papers)·Rakshit S TrivediNatasha JaquesLogan CrossAlexander Sasha VezhnevetsJoel Z Leibo

8 days ago