Gary Sutton

And even if you want to count Thomas Bayes and conditional probabilities centuries ago.

So AI is empowered by, from an algorithm perspective, it's not powered by anything new.

It's using algorithms that were developed many, many years ago by originators who had no idea how their creations would eventually be used or used in this way.

That's half of it.

The other half is, yeah, those same algorithms are being applied at scale thanks to engineering breakthroughs.

OpenAI demonstrated just a few years ago that you can have more predictive models by training your models on larger data sets and throwing computing power to be able to churn through that data effectively and efficiently.

So I like to think of it as brain power versus processing power.

Yeah.

It's...

So it's not the brainpower that is further powering modern AI.

It's really the old brainpower, I guess.

The old algorithms with the processing power, which is why you've got these massive data centers that are being constructed.

Well, it goes back to the findings from OpenAI.

It's not to get better performing models.

It's not uncovering other algorithms or generating new ones.

It's using the same algorithms, but now applying them at a much, much larger scale.

So train them on larger data sets, so larger volumes of data, more parameters, more processing power, more hardware to churn through that data.

that's really where you get the performance differences and the performance gains.

So that's why I say it's more processing power than brain power because we're just throwing more hardware at it.

But there's value in doing that.