TagLast edit: 26 Nov 2020 11:23 UTC by Remmelt

When some humans want something done, those humans can delegate responsibility for the task to one or more AI systems. From the perspective of the AI systems, the relationship would be one of assistance directed toward the humans. However, to avoid dependence of our arguments upon viewing AI systems as having a “perspective”, we treat humans as the primary seat of agency, and view the humans as engaged in delegation.
– excerpt from AI Research Considerations for Human Existential Safety report

Del­e­gated agents in prac­tice: How com­pa­nies might end up sel­l­ing AI ser­vices that act on be­half of con­sumers and coal­i­tions, and what this im­plies for safety research

Remmelt26 Nov 2020 11:17 UTC
7 points
5 comments4 min readLW link


lsusr29 Oct 2021 7:29 UTC
30 points
4 comments1 min readLW link

Map­ping the Con­cep­tual Ter­ri­tory in AI Ex­is­ten­tial Safety and Alignment

jbkjr12 Feb 2021 7:55 UTC
15 points
0 comments26 min readLW link
No comments.