Lex Fridman: belief

Transcript context

…Yeah, that seems like a thing that you could… I could definitely encourage the model to do that. I think it’s interesting, because there’s a lot of things in models that… It’s funny where there are some behaviors where you might not quite like the default, but then the thing I’ll often say to people is, “You don’t realize how much you will hate it if I nudge it too much in the other direction.” You get this a little bit with correction. The models accept correction from you, probably a little bit too much right now. It’ll push back if you say, “No, Paris isn’t the capital of France.” But really, things that I think that the model’s fairly confident in, you can still sometimes get it to retract by saying it’s wrong. At the same time, if you train models to not do that and then you are correct about a thing and you correct it and it pushes back against you and is like, “No, you’re wrong.”, it’s hard to describe, that’s so much more annoying. So, it’s a lot of little annoyances versus one big annoyance.We often compare it with the perfect. And then I’m like, “Remember, these models aren’t perfect, and so if you nudge it in the other direction, you’re changing the kind of errors it’s going to make. So, think about which are the kinds of errors you like or don’t like.” In cases like apologeticness, I don’t want to nudge it too much in the direction of almost bluntness, because I imagine when it makes errors, it’s going to make errors in the direction of being rude. Whereas, at least with apologeticness you’re like, oh, okay, I don’t like it that much, but at the same time, it’s not being mean to people. And actually, the time that you undeservedly have a model be mean to you, you’ll probably like that a lot less than you mildly dislike the apology. It’s one of those things where I do want it to get better, but also while remaining aware of the fact that there’s errors on the other side that are possibly worse. I think that matters very much in the personality of the human. I think there’s a bunch of humans that just won’t respect the model at all if it’s super polite, and there’s some humans that’ll get very hurt if the model’s mean. Yeah.…

Stored transcript either side of the excerpt. The highlighted words are the published quote; the surrounding text is unedited source, never generated.