Tim Davis

I hold concern there, I guess I would say.

I do think perhaps we need a different approach.

And I'm not in any way suggesting that I know what that approach is.

But I often wonder, are we on the right vector?

And humans have a nature of just trying to go to the simplest path and things.

And sometimes I think we do need some more contrarian thinking in innovation, particularly in this space.

You know, I think what is likely, and I know, you know, even back at Google, I think this was a thesis that we had, you know, back in 2017, 2018, was...

Look, at the end of the day, distillation is now an incredible process.

This was something that had been invented earlier on.

You have these massive models and they have this incredible history of human knowledge contained in them.

But again, if you just take that and you put it out into the real world, it's not like it has an understanding of an object if it's occluded, that it continues around the entirety of the object.

Just like when children are first born, they recognize physical concepts pretty quick.

They don't need to be trained on some historical corpus of data.

Is it like a CPU?

I think it has to be more powerful than a CPU.

I think there has to be some generality to the infrastructure, no question.

But if you look at where we are today in AI in some ways with GPU capacity and how fast it's been scaling, I sort of in some ways believe it's reminiscent of the first sort of

foray into cloud infrastructure.

There was one major provider at the time that was AWS and everyone, you know, first, you know, traditional enterprise was like, ah, we would never go to the cloud.

You would be crazy to put your data in the cloud.