Rob, Luisa, and the 80000 Hours team

Yeah.

It is worth having some people thinking about stuff more than 10 years out, given how impactful some of these changes could be.

Exactly, exactly.

Yeah, how can, I guess, people who do want to, maybe they think you've said the wrong thing, but they want to say the right thing.

How can they go and, I guess, get involved in this debate?

Yeah.

If I imagine someone who would say that this isn't really useful work, I could imagine them responding that there's so many things going on.

Like, this is the most difficult sort of futurism, the most difficult social science you could imagine.

Because you imagine, like, many fundamental assumptions about the world have changed.

We're not, like, sure which ones are going to change and when they're going to change.

And then we have to, like, we can barely even understand that what exists now.

We don't even know, like, necessarily why do we have the government structures that we do now, let alone what would they be in future in some different condition.

Yeah, Carl Schulman proposed this on the show a year and a half ago or something like that, I think.

So I'm so glad to see that it's actually going ahead.

Because it's very difficult to avoid data poisoning, right?

You're saying, well, we want to train like ChatGPT 1950 that only has access to any text that came up until 1950.

Well, one challenge is there might not be enough text.

It might not be a very smart model.

I guess another thing is how do you avoid any knowledge about events that happen later kind of sneaking back into the text unrecognized?

Yeah, it's a bit of a clue about how the future might have gone.