Can AI be slowed down? Stanford HAI opens its Prompt Response series
Stanford HAI says the opening session examined emergent agent behavior, recursive self-improvement and whether a kill switch can make advanced AI safer.
TLDR
Stanford HAI describes Prompt Response as a new series offering independent, evidence-based critique of contested AI questions. It says the first session explored whether AI can be slowed down, looking at agent behavior, recursive self-improvement—AI repeatedly improving itself—and whether a kill switch could improve advanced AI safety.
Combined views
5.3K
2 Sources, first seen ago