Nathan Lambert: uncertainty

Transcript context

…Taking it as something silly. Something as silly as the spelling, which Brits and Americans will laugh about probably, right? I don’t think we care that much, but some people will. But this can boil down into very, very important topics like, hey, subverting people, chatbots, right? Character AI has shown that they can talk to kids or adults, and people will feel a certain way, and that’s unintentional alignment. But what happens when there’s intentional alignment deep down on the open-source standard, it’s a back door today for Linux that we discover or some encryption system. Chinese uses different encryption than NIST defines, the US NIST, because there’s clearly… At least they think there’s back doors in it. What happens when the models are back doors not just to computer systems but to our minds? Yeah, they’re cultural black doors. The thing that amplifies the relevance of culture with language models is that we are used to this mode of interacting with people in back and forth conversation. And we now have very powerful computer system that slots into a social context that we’re used to, which makes people very… We don’t know the extent that which people can be impacted by that. So, this is an actual concern with a Chinese company that is providing open-weights models, is that there could be some secret Chinese government requirement for these models to have a certain back door. To have some kind of thing where-…

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