Dax Shepard

And so in that sense, it's an applied problem.

The body has simplified it quite a bit by the time we get to the wrist.

That's right.

The challenges of it are, how do you get something to work on every single person when they put it on the first time?

That's a really hard application problem.

Because it has been determined that people won't have the tolerance to let it learn from it for some period.

The consumer doesn't want a learning phase.

Yeah, we call that personalization.

So we, so far, have not gone down that road.

We just have a general model.

You put it on and it works.

And what that means is you have to collect quite a bit of data.

We wrote a paper, actually published it in Nature, about some of this work.

And so what we learned was that if you get a bunch of people come in and you have them do the gestures and you machine learn the hell out of it, then with a lot of effort and clever engineering, you can make it work.

So what probability, like what percentage would you say it's accurate?

For the things that we've released, it's very, very accurate.

In the 99 something?

Oh, yes.

I don't actually know what the numbers are, but they're very high because you want to have something out there that's quality.

Yeah.