Alright then, basically what I taught you already was that an agent is basically anything

that has some sort of way of perceiving the world, some sort of way of interacting with

the world and of course if we do a formal definition we also have to include this same

world in some way. So an agent usually acts in an environment and it perceives it via

sensors. Sensors is a very general term for GPS signals or eyes or your skin tells you

how warm it is or whatever you want really. Anything that gives you any information about

the environment can be classified as a sensor and agents also have actuators that are things

that can change the environment in any sort of way. For example turn down the thermostat

or speak something into the room so people know what you're talking about or throw a

pillow or anything. Any sort of that, anything that can interact with the environment and

change the environment in some way is classified as an actuator. Which brings us to this wonderful

diagram. The environment is being perceived via the sensors of the agent and the agent

based on these percepts does some actions and changes the environment and this goes

on in a loop basically. Some examples of agents are humans, robots, soft bots, so software

bots, thermostats in a sense and else. Can anybody to avoid you all falling asleep, can

anybody give me an example for something that is not an agent? I'm pretty sure you can.

Overhead projector. Why is it not an agent? Because it has no sensors. Yeah. I hope so.

And it doesn't act independently. Independently, yes. That's actually quite a good answer because

in a way we are usually, we could say okay the overhead projector has some sort of a

sensor. Cable goes into my laptop and then goes into the overhead projector so the overhead

projector senses whatever input is coming from my laptop and it acts as in it displays

the thing on the thing. But that's not really the definitions of sensing and acting that

we usually take. But of course it's in a way complicated and a little bit philosophical

what we can categorize as an action in this case. But usually we want to have some sort

of independence in there and we have to some sort, not just a mechanical reaction, just

display whatever is there or purely react on input. But we usually want to have some

sort of deliberation independent action in there to call it an agent proper. You seem

skeptical?

Actually, you should talk into this.

Projector may get too hot, it might turn itself off and so it acts autonomously.

Independently, yes. Yeah, that would be a way of recategorizing the whole thing. Then

we also have the sensor of internal temperature or something and then we have a percept history

of both the, mostly of the internal temperature and then you could say, oh well, if the temperature

is above this then shut yourself off or if not then just mechanically display the thing

that's coming in via input. I think that would be fair to classify as an agent at that point.

But if the projector just only displays whatever is in there, that's usually not classified

as an agent. Does that make sense? Another example for something that is not an agent.

A chair?

A chair, yes, absolutely, not an agent. More? Okay, I'm not trying. Would you give that

over there?

I would say that the classical industrial robot is not an agent because it has no sensors.

It's just an arm doing a movement all the time.

What is not an?

A classical industrial robot. Like a robot arm that is just moving and I put my program

inside and it has no sensors looking at the environment. If there is a human coming inside,

it just moves and continues and.

Yeah, that works. If you program an industrial arm or something to just do this motion over

and over again, then it might be very helpful for you because it maybe picks something up

there and puts it on another conveyor belt or something, but it doesn't really perceive

anything there and acts on that. It just does the same thing over and over again. Yes, question?