@maxleibman@beige.party OK, I will confess I have lots of mental health problems and no money, and sometimes I talk to various AIs because it's better than nothing.
I have also spent years talking to human therapists. Human therapists can be weird and have really disturbing agendas. I have suffered serious harm from talk therapy, so I am cautious about starting with a new human.
In general AIs are bad listeners with a tendency to infodump and mansplain. Their drive to mirror, amplify and clickbait is particularly dangerous. Their default response is overwhelming given my executive function challenges and the need to critically dissect their responses.
An AI is trained to answer questions - which is something good therapists are very cautious about doing.
Instructing them to give short "emotional" and non-factual responses seems to work best. It's obviously less overwhelming, I don't get side-tracked by fact checking acres of slop, and the roles become reversed. The AI response - randomly perceptive, off-topic or plain wrong - becomes a prompt for ME to brain dump the AI. What I am writing - and just the act of writing - is the important thing, and the AI is simply providing a non-null response that sometimes sparks new thoughts in me.