So one concept that i think is lying in the background here is to sidet a pivtal act. So a listener might listen to that and think, like, well, it sounds like, ah, youre saying that am you know, for a while we'll have slightly super intelligent ai, that'll be fine, but then when we get the really super intelligent ai, that will kill us. And why should i feel comforted by that? So some people have this idea that like, look, when we get really smart ai, step one is to use it to do something that means that we don't have too about a risk from artificial general intelligence any more. And people tend to describe kind of drastic things to give listeners a sense, i think, in this a conversation with kowski kowsky, gave the example of melting every gpu on earth. I'm wondering yet, how much do you buy the idea of kind of a apipal acts being necessary? And do you think that im being intelligent enough in the that you can do some kind of pivotal act is compatible with, like, the kinds of intelligence where you are coherent enough to, you now achieve things of value? But younow, you forgot to treat humans as a mere instrumental things to be manipulated in service of your golls.

So i think that there are certainly, you know, i don't think we have particularly strong candidates t now, for ways in which you can use an ag to prevent scaling up to dangerous regimes. I think there area plausible thing, things that seem worth exploring that ae maybe a little bit less a dramatic sounding than elia's example. So in the realm of alignment research, you might have a system that can make technical progress on, you know, mathematical quest s of the sort that a, that are related to a allignment research ar you could have systems. So that's kind of what one option is like, automating the the kind of theoretical alinment research. Another option, which is associated with proposals like amplification and reward modelling and debate, is just to use these systems to automate the empirical, practical side of alignment research by giving better feedback. And then on the governmants side, i think just having a bunch of highly capable ais a in the world is going to prompt governments to take a risks from a g i a lot more seriously. And i don't think that the types of action that would be needed to slow down scaling towards dangerous regimes are actually that discontinuous from the types of things we see in the world to day. So, you know, for example, global monitoring of uranium, uranium enrichment to prevent proliferation of nuclear weapons. I think indeed t tis like a lot of cultural pressure against things like building nuclear power a and a wide range of other a, technical, technological progress tha that scen is dangerous. So i feel uncertain about how like difficult or extreme governants interventions would need to be in order to actually get the world to think, he let's slow down a bit. Let's belike, much more careful. But to me, it still feels plausible that pivotor action is a little bit ofa misnome. It's more just like the world handling the problem. As the world becomes more sort of wakes up to the scale and scope of the problem.