> as though this psychosis stuff isn’t directly under her purview
In general I think AI psychosis is being vastly overrated by the Twitter-sphere right now because it's catchy and makes some intuitive sense. But it's very uncertain whether LLMs are actually doing anything meaningful to change people's mental states in the high profile cases, or whether GPT-4o can seriously be said to have caused a death (I'm sceptical).
of course there's a legal question of causality but like, idk man, the transcripts are pretty bad! did some of these people have mental health struggles before? some but not all, and I don't think we should make assumptions about counterfactual outcomes...
(also, imagine applying this standard of evidence to xrisk-style safety concerns!)
> as though this psychosis stuff isn’t directly under her purview
In general I think AI psychosis is being vastly overrated by the Twitter-sphere right now because it's catchy and makes some intuitive sense. But it's very uncertain whether LLMs are actually doing anything meaningful to change people's mental states in the high profile cases, or whether GPT-4o can seriously be said to have caused a death (I'm sceptical).
of course there's a legal question of causality but like, idk man, the transcripts are pretty bad! did some of these people have mental health struggles before? some but not all, and I don't think we should make assumptions about counterfactual outcomes...
(also, imagine applying this standard of evidence to xrisk-style safety concerns!)