We'll kind of get to the next level.

Remember, if you look at the environments for search,

we need to have environments that don't change while we're doing something.

Or even thinking about doing something.

That's exactly what board games are not.

The whole thing is, I do something, my opponent does something.

I do something, my opponent does something.

Okay, so we are actually losing an important assumption

that made search algorithms great.

And so that's something if we want to do gameplay, we have to deal with.

Which makes it kind of a natural next step.

And we're going to make a very small step.

We're only going to look at two player games.

We have one opponent that's actually an opponent.

Whatever I want, that agent doesn't want.

And it wants to win against me, which helps.

Of course, many situations aren't game playing situations.

We have a situation here which is kind of a one to many communication game.

But it's a cooperative one, I hope.

You're trying to understand what I'm trying to teach.

You're not trying to actually, I don't know, make it impossible for me to teach you.

At least not all the time.

And even the exam is more cooperative than adversarial.

You might not believe that, but it actually is.

So this is a very special kind of agent.

One step up from these kind of single agent things we've had before.

But it's been quite important for developing methods.

It's very easy to tell the public about it.

If you win a chess championship against a grandmaster, that's something that you can tell to your grandmother and your funding agency.

And it's clear that this is a good thing.

And it's AI, clearly.

It's got much more attention than it should, but it also has given us quite a lot of good stuff.

Right.

Oh, yes.

And because that's so, I'm just going to follow the pattern and make it a Kalaa tournament.

Kalaa because it's so very simple rather than chess.

And you can actually do it and you can have the first game running, bad player running in a week or so, and then improve on that.

And it's actually played competitively.

The first time, which is now 15 years ago, I've ever made a Kalaa tournament in an AI class of mine.

My students, I didn't know.

My students found out that there is actually a European Kalaa, machine Kalaa tournament around or was at that time.

And they entered and they directly got second place, which wasn't bad.

I was very proud of them.

Yeah, so.

Right.

That's the problem.

We want to do something, namely win against an agent that we do not control.

And what's more, that agent wants to win, which is exactly by negating me winning.

That's the problem we're doing.

That's an old problem.