Good. So, I'll try to stop in time.

We've looked at planning.

We haven't really looked at good algorithms,

but the theory tells us the algorithms will involve heuristics.

Because complexity tells us it's going to be

exponential without good heuristics.

So, what we're going to do tomorrow is we're going to look at general purpose heuristics.

Just like we did for general search problems,

we will look at heuristics that are essentially solving the problem with cheating.

Or in a more polite way,

saying we solve relaxed problems and then kind of lift the solutions.

We're going to go into quite a lot of detail there and see what one can do.

With that, yes.

So, we've looked at strips first,

which is essentially the idea of using language level,

action descriptions with add and delete lists and preconditions.

We've gotten tired of using propositional logic because we are humans and we don't think in propositional logic,

but we like our words every and for all,

even if they only range over a finite amount.

It makes it easier to describe the world.

All the AI1 students are at least in their fifth semester.

It's much better of saying this guy fifth semester,

that guy seventh semester, that guy ninth semester, and so on.

Much more work.

I can say all of them at least fifth semester at most ninth,

which is something we could read off the evaluation.

That's shorter. Then the machine can go and say,

this guy fifth, this guy at least fifth,

that guy at least fifth, and so on for all the 120 that are still in the course.

Or no, in this case, 21 who've actually answered the evaluation.

I shouldn't be making assumptions about the rest.

Which is why we came up with a more practical language.

That language is so practical that it's actually used in industry.