Hello. In today's video nugget, I would like to briefly explain about other ways of doing

knowledge representation, which I'm not going to do in this AI course. Remember, we were

still in the introduction of the area of knowledge representation. We've seen two examples. We've

seen semantic networks and the ideas behind the semantic web. Semantic networks basically

focused on T-boxes, how certain concepts relate to each other, and

the semantic web, which really is all about a giant A-box of facts, or web of data, as they say,

where the T-boxes, the ontologies, actually play the role of giving meaning to the concepts,

and thereby giving them a process model. There are other things one might want to do when we're

doing knowledge representation, which I would like to go into now. So, if we look at the predicate

logic way of writing down knowledge about the world, we've already talked about this notion of

locality. That is something where, actually, I claimed we're losing when we write down things

in logic, be it predicate logic or anything else. That's something that people noticed, and they

came up with a notion of so-called frames, where we're kind of making the whole thing more local

by saying we have an object, in this case, maybe a cache object, which has an instance, cache 22,

which basically has a couple of predefined slots, which have values. The totality of

the slots and values actually give you the meaning of cache 22. You can directly see that

you're just redistributing the information here. The thing is, we're using cache 22,

which supposedly is the object, the cache object, which has a classification and two relations.

We're basically taking the subject here and its classification and just basically say it once and

get to a, many people think, more attractive and more localized version of the whole thing.

Frames were quite influential in the development of both knowledge representation and AI, but also

they had some influence on what we now call object-oriented programming. Then, of course,

there is the notion of time. You think about the things we've talked about in semantic networks,

mostly static stuff. We have knowledge about animals and whether they have heads and legs

and stripes or something like this. Very much of our knowledge about the world really involves time.

The thing you typically do when you get your haircut at a beauty parlor is kind of structured

like this. You make an appointment, you go there, you talk to the receptionist, then you get your

haircut and, of course, that's an old big process in itself. Then you pay and if you're happy,

this is the standard example from people in California, you give a big tip and if you're

unhappy, you get a small tip. In Germany, this part might be a little bit different.

Here we have a notion of time and eventuality, which is captured in something called a script,

which is really a way of representing the knowledge about how certain eventscapes actually take place

and what their order is. The order is the important thing here. You must make an appointment before

you go, before your hair is cut. Otherwise, if you don't have an appointment, you can go there, but

you won't be... There's a big chance that you're not actually going to have your haircut.

This is also something that tells you something about the world. Very often, this is

used to understand how the world works and how stories work. For instance, you

can say, I got my haircut, the receptionist, blah, blah, blah. You can ask yourself,

the receptionist basically tells you there is a receptionist of which without this script,

you don't know anything. You even have this the that tells you there's exactly one receptionist,

which you can... If you know the information in a script, then you can actually resolve those

things that would otherwise be un-understandable. This is another thing of knowledge representation,

which we're not going to go into. Then there are more other things like there are procedural

representations in production systems where you basically think of the world as being governed by

a couple of rules that generate all possible behaviors. You may think of Conway's Game of Life.

You have analogical representation, which I'm not going to cover. There's something called iconic

representations. The things that you might see, the knowledge representation, when you buy a new

coffee machine where you have all these funny little drawings, or if you go to IKEA and want to

assemble the furniture you've just bought, then the knowledge about how to do that is actually

given by little drawings, which is very nice, which has lots of interesting research questions for AI,