NickRiviera
8mo ago
They also tend to always say that everything will be ok; turns out everything did not turn out ok. I could have bought an awesome drone for the money I spent getting lied to and it would have brought me more happiness long-term.
Makes me wonder if you're the Scott Alexander alt because this is clearly a mental health practitioner's opinion. All LLMs go off the rails if you keep talking to them long enough, that's a technical problem to be solved in the next year or two, not a reason that human therapists should have jobs ten years from now. OpenAI has already made it a non-issue by just limiting ChatGPT's context window, you'll see this issue more on models that let you flood the context window until the output quality drops to nothing.
Just FYI, a lot of people would much rather spill their guts to an AI than to another human. Also, one of the most common kinds of stress people face is financial stress, and for these people paying for a therapist will cause more stress than it will ever resolve. Mental health professionals are much more useful to the people that need them most when they are free. Far more people will kill themselves due to not getting expensive human attention than will ever kill themselves because their cybertherapist told them to.
Even Parisians are no less friendly than the inhabitants of any comparable big city.
This is totally wrong, have you seen how Parisians park? They literally ram the car in front and behind them with their bumpers to make room for parallel parking. Strangers will scratch up your bumper to make their parking easier and won't see anything wrong with it. This would be considered extremely rude in most other places.
Here's a video of Americans being shocked by Parisian parking: https://youtube.com/watch?v=n51OdFlOi1o
Have the Therapist-LLM write post-interview notes like a psychiatrist would. You could even come up with a code to signal a red-flag status to the IRL therapist in a way that would be invisible to the patient reading their own note. (e.g. it being written in italics or all-caps could mean that the red-flag detector was triggered during this session)
More options
Context Copy link