Peter McCrory

Now, to your point about

the facts that these models are getting better, one reason to think that this is an underestimate is it is current usage of current generation models.

As businesses figure out how to adopt these tools,

much more scalably and efficiently for a wider range of tasks in the economy, that would push the number up.

And for roles that are more at risk of automation, I think there's a greater risk of displacement.

Add to that, the models themselves are just getting better.

And so that's like one way to, this is a very large number, and there are reasons to think that it could very well be

much larger in the future.

So I would say that it's generally consistent with this point that I like to make, that

adoption of AI has been much faster than previous technologies, consequential technologies like the rollout of the internet or the personal computer.

You know, hitting rates of adoption in two years that it took those technologies five years, and certainly much faster than the rollout of electricity, which required a build-out of infrastructure.

Now we see

other evidence of very fast adoption in the Anthropic Economic Index.

So we look at differences in adoption rates across U.S.

states.

And what we see is that states that are late adopters, so there's like low overall usage, they're actually catching up much faster to early adopting places like New York and California.

about five to 10 times faster than that diffusion process typically takes.

That's based on a reference to some academic work that was published last year.

So I think this is an internally coherent story that not only is it a broadly applicable technology, but people and businesses are very quickly finding immense value out of using these tools.

So I don't have much of a perspective on the marketing angle, but what I can tell you is what we see in the data that we produce.