OpenAI AI for Mental Health - OpenAI Needs to Be DEAD

Eli the Computer Guy (0:00)

This story is squarely in the category of just because you can doesn't mean that you should. So this is a hell of a thing. Coming directly from OpenAI itself, introducing Mental Health Bench, an open benchmark developed with more than 80 licensed mental health experts to evaluate AI responses in realistic mental health conversations. No, no, no, no, no, fucking no, fuck, fuck, no, fuck, no.

How do you like that argument, Sam Ullman? I rest my case. No, fuck, no, fuck, no.

Prosecution rests. So, so yeah, I think this is kind of an interesting story just to kind of show you how stupid these frontier labs are, right? So these frontier labs, they want idiotic valuations. Anthropic wants a two trillion dollar valuation. OpenAI is trying to get a 1.5 trillion dollar valuation on the private market, and they're trying to prove to the world why they're worth this amount of money. They're not. They're not. So they're throwing everything against the wall, trying to find something that sticks. So we see this with Anthropic right now. Anthropic came up with their own bio lab. Yeah, so they actually have a wet lab so that their AI systems can think about biology and then physically implement whatever the hell it is that they're thinking because that's going to make the world a better place or kill humanity. I just do want to put this on the table that Dario started talking about AI, like wiping out humanity. And then we find out that Anthropic has a bio lab.

Like, oh, oh, yeah, that does change the conversation just a wee bit there. Anyways, now we have Sam Altman, right? Sam Altman comes out. And again, there's all these discussions about how AI is going to destroy civilization. And what we've heard from OpenAI is that people are using it for therapy services. They're using it for health services. They're using it for all of these different things. And one of the questions that I keep bringing up is fucking why? Why do you allow this, right? You think about this with any kind of technology system that you're going to build, there's going to be a purpose. There's going to be a purpose for the system that you put out there, and then users are going to do whatever the hell users do. And many times, what you do is you create guard, you create old fashioned guard rails.

Basically, just to get the users to use your system the way the hell that they're supposed to, right? You know, basically, if they go out of bounds, essentially, you don't allow them to do that. I think what is interesting here is we know that there is a real problem with people trying to use AI for mental health services, right? We know it has led to the deaths of people. It has led to the suicides with people. We know it has created great and massive harm. And so the interesting thing here, right, is if these AI frontier labs were halfway sane, if they were not run by the devils incarnate themselves, basically what should happen is very primitive guardrails should be put in place on the platforms that simply don't answer those questions. How do I deal with depression?

Answer. Great question. Don't fucking ask me about it.

Right? That is basically what OpenAI or Anthropic should push back. But here's the problem. They need more usage. They need more users. They need more usage. They can't tell anybody no, because if they tell anybody no for any reason, then they're going to lose that user, that they're going to lose that usage, and then their stats are going to fall. So come hell or high water, not just the AI, but the lab itself, you got to give an answer. Again, one of the interesting things with hallucinations is that you do get an answer. Like the model doesn't say, I don't know.

How easy would that be? I don't know. The model almost never says, I don't know. The model will always give you an answer, whatever the answer may be. And that's one of the things right now, it's like with these AI Frontier Labs, right? They feel, they feel, again, company culture, they feel that they have to give you an answer. They have to provide services, even when it is just functionally a bad idea. And so now we have Sam Altman, and we have OpenAI, and so OpenAI is coming out and saying, well, since people are using our services for mental health, instead of just putting up rather primitive guard rails to say, fuck off, no. Now, we're going to start scientifically looking into this more, to understand better how AI can deal with mental health issues. People turn to AI for many kinds of conversations, navigating a difficult relationship, working through everyday stress, supporting someone they care about, or deciding how to approach a challenging situation. These conversations require accuracy, practical judgment, and respect for people's agency.