Dr. Richard Moulange

So the best human teams get something like 40% on the eval, which is still lower than the state of the art from AI systems.

It didn't persuade everyone, however, and what really worries me here is that I think it's partly that people just didn't know it happened.

I still read, you know, in newspapers, in op-eds, and also meet people at conferences who are often experts in sort of maybe biosecurity in general or in security studies, but don't deeply follow the AI angle, who say, oh, but yeah, this tacit knowledge thing is just, you know, it's a huge barrier.

We'll never overcome it.

And I say, oh, what about, you know, virology's capabilities test?

Don't you think secure bio, you know, really provided evidence that sort of questions that?

And they're like, oh, what's that?

I've never heard of it.

And that's my larger problem, that I think some of the best work is not being, I don't know, it's not getting around.

It's not being disseminated.

I don't want to blame anyone in particular.

It's just unfortunate it's not happening.

And I worry that decision makers do not have necessarily all the sort of lines of evidence that they need to be able to make an informed decision.

And that this actually compounds where they will have expert advice that is misinformed because those experts actually aren't keeping track.

Yeah.

We don't really know how much AI is uplifting beneficial life science.

To my knowledge, there isn't sort of a deep, regular survey to understand how AI might be helping top scientists.

I think this is a really important thing.

It's important not just because it might proxy how AI could uplift malicious actors.

It's important because genuinely we want to know how AI can uplift beneficial science to improve public health, to generate new drugs, to advance human progress.