Sam Parr

And they're like, but the BlackBerry, you know, keyboard, like you gotta write emails.

He's like, no keyboard on the phone.

And they're like, but how would we, the accuracy, I mean, screens today don't, he's like,

no keyboard on the phone.

And so then they had to go invent multi-touch and figure it out.

So he gave them the constraint that you got to do it in this.

These are the constraints, but then I give you the time to go explore and figure out like which path might work.

That's interesting.

And they, they also did this with the game thing, by the way, when they did the go thing, the first one was again, trained on a hundred thousand games.

Then they created alpha zero where they said, now try to make it win with no prior human knowledge.

Because he's like, if we're ever going to do new novel things, you got to assume we're not going to have a database of 100,000 good humans at doing this to use.

And so they did.

They created AlphaZero, which could win in chess and go just by playing itself like 10 million times or whatever.

It figured it out.

And so similarly here, they were like, you got to go back to the drawing board.

And he described, he goes, first, I'm going to give them the constraints.

Second, I'm going to let them be creative and try to go to the drawing board, figure out multiple different possible ways this might work.

He goes, and then when they pick one, he goes, I know this is when it's time to push.

He goes, because first we will get worse than we were before.

Then after some time, we'll pick an approach and we'll get right back close to where we were before.