Time's up.

Let's go back to game playing.

Remember we're in a sit... yes?

Because we only had seven questions.

You would be surprised how difficult it is to come up with halfway decent questions and

sometimes we don't manage.

So I usually, unless there's something where you have to do a lot of mental arithmetic

or so or closes which are kind of these texts with choices which take longer, usually I

take about one minute per...

I use one minute per question which usually is relatively accurate.

There was however a relatively slow take up today.

I wonder why.

We can see here that the response rates are very low in the first almost two minutes.

I wonder was that because the system was slow or was that because you were slow?

Okay, I have no idea why that might be but... yes?

Oh, how nice.

Are you using a recent Chrome?

Okay it should be SVG and Chrome should be able to do it.

I don't know why that is.

Yeah, okay.

Send me email about it.

Okay so but one of the things that is speed is a bit of an issue.

Speed because in the exam there will be time pressure.

Those of you who've been in an AI exam and that's intentional because learning also has

to do something with availability.

If you just basically...

I always say, well this you could reinvent if you had a week.

Well in the quiz and in the exam.

Or if you're doing work with these things then you don't have a week.

Then you want to be efficient and just know things.

And yes I know there are other courses as well.

I've been a student.

The other courses are just less important.

Okay take that with a kilo of salt.

Righto we were talking about games and games we're going to look at as search problems

only that this search problem is slightly different because we have an opponent and

we could think of this as a non-deterministic environment but we can also think of this

as a two agent in this case.

Situation there's this other guy and that other guy wants exactly what we don't because

we restrict ourselves to zero sum games.

So we adapt the math accordingly and we kind of split the search problem into the half,

into the max stuff and the min stuff and the max actions go into min states and the min

actions go into max states and kind of the ping pong between the two agents is nicely

reflected here.

So it's reflected in kind of the terminology where any move max makes is actually half

a move even though we've called it an action before.

Okay this almost takes a while.

The next thing we talked about was even though mathematically we're looking for a strategy,

something that always in every game state tells max what to do and correspondingly min

but these things are computationally completely infeasible to many states.