David Duvenaud

I guess I would say there was no slam dunks.

There was sort of, to me, the beginnings of a field.

I mean, the coolest thing that happened, there was two cool things that happened, I think, because of this conference in particular.

So one was

we invited one political science student who worked with Alain Defoe to talk about how competition between states in theory and in practice puts constraints on how much welfare they can spend on their citizens.

And basically it's like an arms race ends up taking money from the poor sort of thing.

And then there was a reply, someone actually applied to the conference saying, oh, in contrast to McInnes et al., we have this sort of game theory model that shows that actually in a bilateral situation where there's only two powers, there is actually a stable equilibrium where they do still end up spending a lot of money on their welfare.

And I was so happy that there was somebody making a position clear enough to be rebutted by someone else.

And this is exactly the sort of thing that I'm hoping will happen more often, like how we provided a venue for it.

Maybe the other concrete thing that was obviously counterfactually good was Jacob Steinhardt.

We invited him to give a talk.

He's the CEO of Transluce, and he told me he didn't actually have an idea when he agreed to give a talk, but he came up with his idea for addressing some of these

dangers in the future.

And his concrete proposal was let's flood the Internet with high quality data showing A.I.

's doing valuable work, but in a morally aligned way.

And so this is kind of like moral fables for A.I.'s.

And if we flood enough of the Internet with this data, then anybody in the future who scrapes the Internet for their own new LLM is going to train something that's basically aligned by default and basically raise the cost of misaligning A.I.'s.

And I don't think this is like that big of a deal in terms of like it being

like fully addressing any problem satisfactorily, but it made me feel like, okay, there is alpha here in that if you push in this direction, we can sort of get people to think hard about this and make like some progress on concrete questions.

That's an improvement on what we have.