So it's getting cold outside.

Yeah, so we have basically reached the final step

kind of in our progression from black box states in search

to factored states, which allowed

us to do both search level algorithms as well as inference

level algorithms and mixtures, of course.

And the last step is structured representations.

And that is a stage where we will use the flexibility

that we can get by designing our own world description

languages.

And we're going to design these languages in a way

you could say that is computer sciency.

We give a formal grammar that tells us

what is in the language and what's not.

We're going to give it a formal semantics,

meaning a mapping into something we already understand.

In this first case, essentially this said true and false.

Because that's what we're really interested about in the world.

Is this sentence true or is it false?

So we're going to start very small with a formal language,

with a world description language called propositional

logic.

Many of you have probably seen it before.

I want to introduce it carefully here again because we,

because on average I don't like how other people introduce

logic.

And then we're going to build on that,

getting more expressive logics.

So the main example we've always been using

is the Wumpus world, which in its particular shape

is completely irrelevant.

The interesting, the important fact about it

is we need some kind of a world description language that

is kind of better than just constraints

to describe the rules of this game.

If there is a pit, then there's a breeze nearby.

Then there's a breeze nearby, and all of those kind of things.

And so we basically started kind of looking at the percepts.

Those are going to be part of the world description.

And we looked at how the agent might do reasoning.

And there are some other things that are just

kind of derived knowledge.

I want you to think about this thing like OK, or visited,

or possible pit.

Those are things that are not perceived by the agent.

If you actually perceive possible pit,

you're already falling.

So those are different parts of the knowledge,

but still talking about the world, which we somehow

want to have in our language.