Tim Davis

You know, you fast forward a decade.

Now it's like, not only do we want our data in the cloud, now we want to be multi-cloud.

And so, I think your analogy is interesting, but I think the actual inferencing process and the speed at which it has to happen and the latency at which meaningful results have to return to the user.

And there's a quoted statistic that we used to use in Google search was,

but every 100 milliseconds of latency, it was 1% on revenue.

And I think that's been validated also by Amazon.

So in the real world, like if you think about deploying, you know, just on a CPU, for example, in an autonomous vehicle, could you imagine flying down a freeway and I don't know, a ball floats out onto the freeway and because the CPU just is not as powerful as a large, you know, parallel, just in terms of flops and processing,

The car will have hit the ball and be another two miles down the road before you get an inference response that says, watch out for the ball.

I think that's the practical reality, though, that any future system, any future form of intelligence needs to be able to have...

And I think we, as humans, have this sort of biological process that we've inherited over generations.

But it is incredible that we are able to process so fast and respond so quickly to things.

And I think computationally, a CPU can't get us there.

So then the question is, in the real world, I don't know if you've been in a Waymo.

Not yet, but I hope to this month.

They are marvels in the sense of, and we can talk about all the different, the sensors and the different approaches on those sensors on those vehicles.

But at the end of the day, you do need a degree of processing power against some battery capacity and thermal thresholds that enable intelligence to respond quickly.

Because everyone I think is optimizing now for a high degree of perceived intelligence at the fastest possible outcome.

And I think it's actually the nature of consumer behavior over time because of internet products that have been created over the last 20 years.

And I think Google is obviously a big leader there, but that's not to say that Amazon and Meta and other organizations haven't done incredible work.

But what we have done is train everyone to get everything instantly.