In the UK, you can access a psychologist for free, and whilst that's no consolation to my American peers, it's furthering the point that America's healthcare system is fundamentally flawed. Technological problems are ideally solved by technology. Social problems aren't always best solved with technology, they're a band-aid.

Well, something being fit-for-purpose is an important consideration. If it tells people that the world would probably be a better place without their broke, uneducated, unwell self and that their life statistically is unlikely to get any better because they're talking to an AI and not a real human being, that's a huge ethical problem, and the LLM is not an appropriate solution.

I can get LLMs to tell me all kinds of dangerous things. The more text you give it, the higher the chance of those safety measures going out the window, especially if that conversation has a lot of back-and-forth over time. When people explain their problems, they use a lot of words and a lot of negative sentiment.

Not talked about this before with anyone, but as a case-in-point, I managed to make an LLM's safety features go completely out the window over me recalling a low-point in my life. I would not recommend that shit in good faith to anybody.