Now, AI is big and when you do research you have to kind of dig in deep and you have to

ask yourselves in this huge area what are you going to do?

In my research group, and one of the things I can offer you is actually doing AI research.

At FAU and so I should tell you about what we do in my research group.

So if you're interested in AI projects, working for us as a student assistant, doing your

PhD with me or something like this, we have money, we can supervise you well and we always

need, so our main, actually our main research tools are human brains.

We can't kind of use AI for that yet.

So we need fresh human brains from time to time, so should yours be one of those?

This is what we do.

We basically come from logic, which is symbolic AI, which is not where the hype is right now

but where kind of this, where you can already do things.

You talk less to the press but more to your, more at conferences.

So what we try to do is to understand how knowledge works.

What is it that knowledge actually is?

How is it organized?

How do we efficiently do it?

How do we verbalize it?

You have knowledge, how do you actually communicate that to other people?

How do you read and get knowledge?

Reasoning systems are kind of at the heart of many symbolic AI systems.

What we're doing mostly is we're using mathematics as our test tube.

We're actually interested in mathematical knowledge.

Why is that?

We claim that this is an extremely good, if limited, area to study knowledge in.

The reason is that mathematics is so easy.

Mathematical objects have very, very, very few properties.

A group is something you can fully characterize, or a finite group you can fully characterize

if you know its order and a couple of more things.

Compare that to a human being.

I cannot fully characterize him by three numbers.

Difficult, math is easy.

Math seems difficult because we want to know everything there is to know about a group.

We have lots of evolutionary experience in looking at objects like that one and making

predictions about them.

Is he going to run out in the next five minutes?

Probably not.

Is this group going to have a commutator that's non-trivial?

Who knows?

We're kind of disadvantaged in math, but knowledge is extremely overt.

Material texts are very helpful if they exist at all.

You can look at knowledge structures very well in math.

Hopefully we can, at some point, graduate from easy objects to more difficult objects

using the same techniques.

We kind of do three things.

One is foundations of mathematics where we try to understand knowledge in mathematics

and build actual logical systems for that.

That is quite involved, very theoretical, and typically not something we let you help

at the bachelor's level because all the things that are at this level we've already done.

We kind of have exhausted the easy problems.