Grant Harvey

But then also, what's the human reaction going to be to them?

It's going to be very interesting.

The analogy I made at the start around models being like a teenager, it shows that the models don't behave as you expect.

It also reminds you that people don't behave as you expect.

I think that's going to be the two things that are going to be very, very interesting.

It's a different thing.

You just need the first trolley problem-esque scenario where the model has to make a decision and there's a big court case and then who knows what's going to happen.

Yeah.

Yeah, people don't perform... I mean, we ran something, God, three years ago, where we were pulling data from a website and having to validate if it was a real location.

And we had people doing it, and they were doing it with a certain accuracy, and we got a model to do it, and they did it with a higher accuracy.

But when the model got it wrong, it got it wrong in weird ways.

Yeah.

And the client didn't like it.

Because if you're like, oh, 95% might say the correct location, and the other 5% of the time might say a different wrong location...

And instead it's 99% you say the correct location and 1% of the time you say banana.

People are going to be like, oh, that's much worse.

Because people are empathetic and people understand human mistake.

Like it's people, man.

They're the fascinating thing about all this.

We'll be in a world and in like, I don't know, some, let's call it 50 years time, people will be like, wow, you drove places?