Okay, so welcome to another week of AI.

I'll let you find a seat first.

So last week

we finished looking at the prerequisites.

All the stuff I expect you to be able to do.

And we started with looking at

artificial agents.

Essentially the idea of an agent, we are going to use that as the central

metaphor

for intelligence

in this course

is that we have

an entity

that can act on an environment

and that can sense the environment

and behave, therefore behave in an environment.

And we

pursued the idea that it would do so

rationally.

Rationally meaning

that it optimizes some kind of a performance measure.

It's essentially survival of the fittest.

You want to optimize some kind of a performance measure.

In evolution

that is

reproduction, right?

That's the performance measure.

The better you reproduce, the fitter you are.

And the other way around, by definition.

We allow ourselves

other performance measures.

But still the idea, this performance measure

has to be optimized

as far as we can.

We've looked at all the things we don't have to do as an agent. We don't have to

be a mission. We don't have to know everything.

We just have to optimize

as far as we can expect

our actions to be optimal.

We don't have to be successful

even though we might like to

because we don't know everything.

So that's rationality.

And we want to look at rational agents and

the biggest upshot that I want you to understand about last week was that

what is rational

depends on

the environment, the characteristics of the environment. We looked at them, right?

Dynamic and static environments,

fully observable or partially observable environments,