So let's kind of look back at AI,

which means in eight minutes we kind of go over 460 slides or so,

which is like playing the minute waltz by Chopin in seven seconds.

Okay, let's try. So what did we do?

So we started out with trying to convince, to look at what AI is.

We tried it with a definition which didn't really help that much, right,

had lots of problems. And then I introduced this framework of intelligent agents.

Okay, agents and they had, we're going to talk about that a little bit longer,

that had this distinction between states and actions and perceptions.

And that's been with us all the time. We were always planning actions or describing the world

state, finding more about the world space, the state by inference, and then doing actions that

also tells us something about the world. Also we started out with Prologue.

And I can tell you that that part of your life might be over,

unless you liked it so much that you want to do AI again.

And we're not going to use Prologue in the next semester.

But I think at some point you might want, might actually value your experience,

because it's different from programming Java and you, I know you've been doing too much of that.

Okay, so we started with problem solving, which is really, really, really mostly about search.

Search with black box states. With all the advantages and the disadvantages this brings,

we've seen two instances, one in the single agent world, think about playing solitaire

or something like this, and game playing, which is in the multi-agent space.

And then we kind of, so then we kind of left the black box world to go into the,

to partly go into the description level with constraints.

We did knowledge and reasoning, which is essentially going to the description level

at various levels of expressiveness in the logics, right?

Very simple things in propositional logic and more interesting things in first-order logic.

And we saw that if we have enough descriptive power and enough search power,

we get things that are universal. Prologue is a Turing complete programming language.

And it really only has a description language, which is a little bit of first-order logic,

and essentially left to right, top to bottom backchaining search as a search,

as a dynamic model, which gives you programming.

And finally, we looked at planning, which kind of solves a couple of problems we get

by going to the description level.

All of that happens in agents, as we kind of came to see in the beginning, right?

And our main question is what do we put in here to get intelligence?

We don't know the answer yet.

We know what to put in here to get an agent that can schedule Bundesliga.

We know what to put in there to have an agent that kills the wumpus.

And we know something what to put in there to actually get a self-employed truck driver.

OK? Those are nice agents. They're partially intelligent in that they actually solve a

couple of problems that your average dog or cat cannot.

Truck driving, I don't know.

But probably scheduling Bundesliga, they can't.

So that's our question, and we're going to pursue that a little bit more.

And the only thing we really did was we graduated from oval agents to square agents.

We still keep the environment outside the agents.

That is a little bit problematic, but makes life easier for us.

And we basically allow us to think of this as a function, a percept to actions function.

And we had a couple of architectures that we talked about.

We had the simple reflex agents, which really, and that's from kind of knowing more about