Okay, can you hear me? Obviously, yes. So, I wish you a very happy and successful year

2023. A bit much. And no. Yeah. Can you still hear me?

Good. Especially since we will have an exam together. And what is it? Five weeks time?

Something like this. No, four. So, it's very close. I would like to start the new year

with the obligatory question. Are there any questions from your side? So, let me just

very briefly talk about the exam. I do not know what time it is. And I have no control

whatsoever over it. We have about 750 students registered for the exam. It seems that I hope

at least everybody who wanted to register is now registered for the exam. And of course,

you can imagine that 750 students wanting to write a single exam is going to make problems

and interesting logistics problems for our administration. They're going to fix that.

But I don't anticipate knowing about the exact time of the exam or then say 10 days in advance.

I'm anticipating that there will be multiple rooms. So, there will be some kind of distribution

of people to rooms. This typically is something we learn about two or three days earlier.

So, and it's not something I can control in any way. Okay? I just have to deal with actually

making it happen. Any problems with the exam date on your part? I've been getting emails

saying could you please schedule the exam at a different time? No, I can't. And if you

imagine what's going to happen, there's a good chance that we'll be taking either the

Mensa, the ball park, and all of the other lecture rooms around here, or the Congress

in Nuremberg. You can imagine that that's beyond what I do. Okay? So, we'll just have

to somehow make it happen. But if there is an exam that you have scheduled at the exact

same time, we can do things like kind of have you do the exam in some other room and we

escort you to the next exam or something like this so that you can't just tell everybody

else what's coming in or receive what's coming in. That we can organize. Right? I'm trying

to make everything possible for you. Maybe even pleasant, but I'm not sure we'll succeed

in that. The exam, you can take a pen, multiple pens just to be safe, something to eat or

drink, your security blanket, and nothing else. No calculators, no smartphones, no cheat

sheets, nothing. Okay? A brain full of AI would be good to take along. Any questions

about this? I'm assuming most of you know that we have all the old exams, some of them

with master solutions. Many of the master solutions are actually being correct. We have

them online. If you are interested in what the exam is going to be like, look at one

of the younger ones that should give you a feel. It's not going to be exactly the same,

obviously. But that should kind of give you a feel for what the exam will be like. Okay.

Questions from your side? Going to be interesting with so many students. I'm going to probably

need a wheelbarrow to transport the exam. Oh, yes, that's something. The exams will

be printed. You're going to get an exam and you're going to write everything onto the

sheets that we give you. Okay? We may or may not give you kind of scratch paper, but whatever

you put onto the scratch paper will not be graded. Okay. Another thing I want to talk

about is you know we have five TAs this semester, which is nowhere near enough. If there's anybody

you know of who is qualified to TA AI2, and some of you may have already taken AI2 and

feel qualified, we would like to offer you a TA's job. You're helping the crowd who's

not TAing who wants to learn AI2, because we want to give you as much feedback as possible.

We wanted to give you much more feedback than you were actually getting this semester, but

five TAs is a very finite workforce and we're doing our best. Just grading 700 exams is

going to be terrible. So bear with us. Okay. So anybody you can, any TAs you can grab and

deposit at my doorstep, wonderful. Okay. Do you have any questions now? Then that's not

the case. Let me try and situate where we are. Well, first thing you're in the AI lecture,

AR1 to be concrete. And that's all about building rational agents. Agents that, given some kind

of a perception of the environment, somehow compute actions or action sequences from percept

sequences, action sequences, and try to do it as good as possible. We have these pictures

where we have the environment, percepts come in from the environment, something weird and