OK, so let's start.

Yesterday, we started the next big block, namely

knowledge and inference.

The idea behind this is that we kind of do

what we learned for constraint satisfaction problems, namely

inference, searching or reasoning

in the space of descriptions and transfer it

to this idea of structured representations.

In structured representations, descriptions, if you will,

are the only thing we have.

We have a language in which we write down

descriptions of the world, assumptions about the world,

something like mammals have four legs or something like this,

or a bicycle has two wheels, all of those things

that are kind of descriptions of possible worlds.

And then we use inference to go from such descriptions

of the world to useful knowledge for agents.

That's what we're doing here.

And we spent a lot of time on understanding our running

example, namely the Wumpa's world, kind of a dark cave

has a couple of problems, i.e.

pits and the Wumpa's, and a couple of opportunities,

namely the gold, and an agent who

has to navigate in this partially visible thing.

And the idea is that the agent actually

keeps a knowledge base, a set of descriptions of the world,

or one description of the world, and uses

percepts and actions to update that.

So just like we had an assignment as being

the description of the world, the partial description

of the world, we now have a language

for describing the world, or what the agent knows

about the world, which is only partial typically.

Conversion of my last lecture is done.

So that's really the example.

We have a couple of percepts, like it stinks,

or it glitters if there's gold, stuff like that.

And I tried to show you that this idea of a knowledge base,

we know things about the world, and our actions,

and our percepts kind of change the description of the world

by incorporating them into the world actually makes sense.

If we looked at these kind of three first things

the agent does, and some of the knowledge here,

that's what I want you to keep an eye open,

some of the knowledge came in via the percepts.

The OK here comes in because the agent is in 1-1,

and notices, oh, I haven't died yet.

And other things come in via the knowledge about the world.

We're incorporating this OK here comes from the fact

that it doesn't stink, and there's no way to know