Okay, welcome to the second but last week of AI.

We're looking at planning.

And the idea is that we want to be able to deal with change,

situations that change in a setting where we have structured world representations.

We've kind of seen over the course of the semester that the more expressive our world description languages are,

the more universal the algorithms become, and the more structure we have to go by to actually guide the algorithm.

In the beginning when we were doing atomic states, we had the problem there are too many of those,

and we can't look into them, and the only thing is to kind of take external heuristics,

which are one-off essentially.

And here we basically need, we have the ability to make, to build heuristics into the algorithm,

because we know about the structure of the states.

And even though we still have lots of states, they are much fewer in practice.

Okay?

Now, one thing that was easy in the general problem solving in Surge chapter was to deal with change.

For any changed state, there was a state.

For any change, there was a new state.

And we didn't care about how many there were.

Now, if we have a structured representation, we have to care about the things that change,

and more importantly, as we've seen in the discussion about the frame axioms, the stuff that doesn't change.

Okay?

Because we can and we have to explicitly represent those things.

So logic-based things actually become difficult.

And planning kind of puts it all into a big pot and starts stirring.

So I'd like to give you a little bit of a history, because it's kind of indicative of what kind of areas like symbolic AI

kind of do in their development.

And basically, this idea of planning is, in fact, a very interesting idea.

It is in kind of the late 60s, early 70s, where after kind of AI being around for 10 years or so,

people were trying to actually get AI into the real world.

And there was a very, very, very influential project at SRI International, the Shaky Project.

And I would like to...

Smartphone connected.

Can you hear this through the microphone?

I couldn't figure out how to hook the computer up, so I'm going to do it in a more analogue way.

And I would like to show you Shaky in action.

So, robot planning and learning.

This is SRI International.

They had a whole floor of the building.

Why aren't you...

Something is not right here.

Let me...

No, I'm going to do it differently.

I'll spare you the music and just narrate it myself.

You can see this, right?

There's a bunch of researchers, since we're in the late 60s, they all look like hippies.

We are in California, after all.

And you see Shaky.

Shaky is a mobile robot with a computer on board and a couple of cameras and a couple of wheels and all of those kind of things.

This computer, by the way, is only the motion control.

It's not the planning.

The planning actually works on a Lisp machine, a computer that actually had Lisp as a programming language.