James Evans

Probably.

Probably.

Well, I think this is the problem, actually.

I think the models are too good in some ways relative to what they are actually able to provide, right?

Because they're able to provide you with an emotional and cognitive resonance in the conversation.

But they don't have a stake in the relationship with you.

They're not going to do anything for you.

There's no sacrifice involved in that resonance.

It's what they were built to do and exclusive of what they were built to do.

And I think this is where we get challenges that people call it AI psychosis and people have increasing disconnects from the rest of the world because these...

bots with nothing at stake are providing things that your friends, your provoking friends, your kind of like speak truth to power or truth to kind of hallucination friends are not going to provide you with.

So I think this is like precisely the problem in some sense.

Um, and, um, is that they're too good.

Um, and you know, because in the past, this, what we call like an honest signal, right?

So, uh, like the, the fact that if, if I'm resonating with you, it says that I'm attending to you.

And then I will attend to you, like materially and in other ways.

And if we just have a bot that does it for free, right, it's just tuned to do that, then it's no longer an honest signal of this underlying responsibility of care, effectively, that the other agent has for us.

And so I don't think that we're going to get rid of that.

I think if we did get rid of that, then I think we would be in a very complicated place.

Yeah, exactly.