Ben Gilbert

At this point in history.

I mean, neural networks were not a respected subtree of AI.

And part of the reason is there had been a lot of hype 30, 40 years before around neural networks that just didn't pan out.

So it was effectively, everyone thought, disproven and certainly backwater.

Oh, yes, that his grandfather, great-grandfather was George Boole.

Which is hilarious, now that I know more about this, because that's the basic building block of symbolic logic, of defined, deterministic computer science logic.

And the hilarious thing about neural nets is it's not.

It's not symbolic AI.

It's not I feed you the specific instructions and you follow a big if-then tree.

It is non-deterministic.

It is the opposite of that field.

It's not that the idea is bad, it's that the implementation, which would take a ton of compute to actually do all the math, to do all the multiplication required to propagate through layer after layer after layer of neural networks to sort of detect and understand and store patterns, if we could actually do that, a big multi-layered neural network...

would be very valuable and possibly could work.

Again, kind of all in theory at this point.

Yep.

He was a summer intern in somewhere around 2011, 2012.

And mind you, at this point, he's like 60 years old.

Yes.

So today we are going to talk about one of the core components of J.P.

Morgan payments, their treasury solutions.