Rob Wiblin

And I guess it's going to be potentially competing with quite a lot of people and quite a lot of compute that is arrayed against it.

We might also have all kinds of different control measures that are constraining the amount of compute that it can access, that mean that it can't work for very long on something before it's being detected.

So it doesn't have as long a leash, potentially.

It doesn't have as access to unlimited resources to try to do these things.

And so each...

Count a measure that you put in place, each extra bit of resilience to make it harder to make new diseases, to try to catch things before they get synthesized, to try to give us more options for tackling a disease once it's released.

Do make it just a less promising project for the AI to engage in at all and maybe makes it not as interested in going rogue in the first place because it doesn't rate its chances of success.

So I would say it's possible.

It is possible that it's overdetermined and that all of this stuff will turn out to have been futile in this project.

But it also could turn out that actually it's extremely relevant and we don't get a super intelligence explosion that moves things out of our hands.

And this will actually make the difference.

So I want to quickly survey what sort of AI biology tools there are that now kind of work and which ones are on the horizon.

So most people will know about AlphaFold, which is this thing that goes from the genetic sequence to understanding the shape of the resulting protein.

And that is like very mature technology now.

We've also got ESM2, which can modify proteins.

So the sequence is different, the amino acid sequence is different, but it's like functionally equivalent.

then one one that is emerging is like beginning beginning to be useful but still has some way to go is like i think protein and mpnn and rf diffusion uh from the baker lab yes yeah where basically you can propose a shape or a kind of target binding that you want a function that you want a protein to serve in terms of a shape and you can get a sequence that um often works yeah an amino acid sequence or a dna sequence that will produce the the protein that you want i think

For some functions, like simple stable structures, that does basically work.

And for enzymes, for catalytic stuff, for, I guess, like more proteins that need to move, it's less so, or it's more touch and go.

We're getting there.