All language subtitles for 15-Lecture 1 Segment 15 Pacman as an Agent.en

So you may have heard of Pacman. Has anybody heard of Pacman?

No, maybe a little bit, okay.

Pacman is a great game from the
eighties. It's a trademark of

Namco Bandai and we're going to be using
it throughout this course to illustrate points in

this educational setting. You can think
of Pacman as an agent.

So, here, you look at this board. That's the Pacman board. What are the percepts?

What do you have when you play Pacman? So essentially Pacman the agent is not the game,

it's not something that draws the walls. What is
it? It's the thing behind the joystick.

What's that?

It's you. So essentially you're the agent when you play a game of Pacman. You're deciding,

do I go up down left right, and so on, and
what is your sensation? So what is the--

for you it involves your eyes, but
basically you see the whole board.

You know where the dots are, you know where the enemies are, you know where the walls are. What actions can you take?

Well you can go up or down or left
or right, and so if we have an agent

which maps the state of the board to
up-down-left-right, the question is,

what goes here in the question mark that lets
it know when to eat a dot, when to run away

from a ghost, when to use the power
pellet. How are these decisions all made?

And you might think, I know, if the
dot's next to me I eat it and if the

ghost is within three squares of me,
run away. And if you try to write this down,

first of all you would end up with the

biggest mess of if-then statements
you've ever seen, and then it wouldn't work.

We will in our projects get a chance
to build agents for games like this, and

you'll see why that kind of thing doesn't work.

Alright, let's take a look at

an agent

for Pacman.

So here is an agent playing the game of Pacman. This is a CS 188

Pacman-like game

and I'll let you watch. As you watch I want you
to think about what kinds of

computation must be going on in this agent in order to be able to

exhibit this range of behavior.

So we're gonna figure out in the coming
classes how we'll be able

to give rise to behavior like that.