Hacker News Reader

Comment by gordonhart | Hacker News Reader

Releasing the weights is an easy and low-cost way for OpenAI to fix this problem.

According to the thesis of this article, releasing the weights would be approximately the worst thing OpenAI could do for these people.

I kind of agree with GP more than the author here. OpenAI got these people hooked and pulling the plug is potentially more harmful than letting them continue to chat with it until they move on naturally (assuming that they eventually will)

I don't think reinforcement is generally the recommended approach for people with delusional or obsessive pathological parasocial fixations. I think generally the idea is to get those people into talk therapy and to cut off all contact with the object of their fixation.

Mere talk therapy is infamously useless for attachment trauma, which is relational and somatic

Begs the question of whether the underlying pathology is attachment trauma or some other delusional fixation! I don't doubt most of the people in this predicament are vulnerable for one reason or another. I made a much more specific claim, about delusional obsessive fixation; I did not claim that their underlying mental health issues could be fixed straightforwardly.

It’s a great question, and I think it’s not exclusive.

Obsessive limerent obsession can be driven by reward circuits, and those _can_ be extinquished by more straight-forward therapy, but if it’s driven by unmet emotional needs, it’s often quickly replaced by some other maladaptive coping mechanism (hopefully slightly less unhealthy).

Reply on news.ycombinator.com