Overall this class was pretty average. Not bad but not good either. I'd still probably take it again but if I had other interesting classes available I'd probably opt for those instead.
In the pro category, Guy is a pretty decent lecturer. He does a good job of going over the points, explaining it clearly, and making sure the class gets it. Only thing I didn't like is he mixes powerpoint slides with drawing on the board, so as someone who likes to just refer to online notes it's a bit of a pain. Also I thought the tests were fine. The midterm was just problems that were covered in homeworks or class. The final was multiple choice, and although I understand the point a lot of the other reviews made here, I thought generally the questions were ok. They mostly wanted you to think critically about the concepts rather than just memorize them all. Also generally speaking the homeworks don't take that long to do, so this wasn't as time consuming as some other CS classes can be.
On the con side, the class is poorly organized, and at least for this quarter I thought the TAs were not good. I stopped going to discussion because my TA seemed to mostly confuse people or not know what they were talking about. They are also pretty poor at responding to stuff like homework questions, scheduling the release and deadlines for the homeworks, grading in a timely fashion, etc. And for the midterm they made you go to a specific TA's office hours to get regrades for each question, so essentially everyone had to go to 3 separate office hours. And in my opinion they graded poorly and would take away points for arbitrary reasons.

I thought this class was very well done! Professor van Den Broeck does an excellent job of explaining the material in an easy to understand way. At the end of the class I thought I had a good introduction to different aspects of AI. The homework and exams were also very fair and very doable. Overall I would definitely recommend this course, I learned a lot and it definitely wasn't a huge amount of work.

The class is OK, professor doesn't use any slide but strictly follow the textbook. As stated by other students, the final is ridiculous, multiple choices, most of them are details and concepts, very little of them require mathematics and logic if any.
The material and topics are somehow outdated, (nearly unchanged for 20+ years http://www.aaai.org/Papers/Symposia/Fall/1994/FS-94-05/FS94-05-001.pdf). It's probablly not be the AI that you have ever imagined. The first half is on searching, the second half covers logic. Till the very end of the class, Bayesian network and machine learning(only decision tree) are included in only 2 or 3 lectures. In my view, the class material is traditional and a little bit outdated compared to nowadays fancy stuff. But they are still good, at least the logic part is intuitive, some material are indeed fundamental of fancier stuff.
For reference, with near average midterm, full points HWs and 10 points above average final I got A.

This class is pretty easy. Lectures are sometimes confusing as we does not elaborate on important topics and breezes by them without clarification. But there are a lot of resources online (ppts on our textbook for example) to study from, so it is not too hard to study for this class. Tests are fairly easy (final was MC, and ppl just finished in half the time and left). Pretty interesting class that could be taught better.

One of my favorite professors so far. He had great, engaging lectures if you actually bothered to listen and follow along. He has a very logical, precise way of explaining things, that melded well with me and I learned a lot from lectures. However, one big drawback is that he doesn't provide notes/slides, so if you don't go to lecture you better hope you can understand the textbook yourself (not very hard till the midterm tbh, but gets progressively harder after that).

In terms of tests, his midterm was challenging, but it was very practical, and if you knew your concepts you could do well pretty easily. The final was however, one of the shittiest CS finals I've ever done. More of a test of your English and your memorization ability than actual CS. About 60% of the test came down to if you had seen that word before and could place it in the textbook. Honestly why would anyone every have a multiple choice exam in CS?

Overall, homeworks, lecture and midterm were fine, but that shitty final dropped me from an A+ to a B+, so I'm pretty bummed about that. Given the choice, I'd probably take the class with him again because I enjoyed his lectures.

He is a nice guy and care about students' learning, but his lectures are not very clear. In fact, this is one of a few classes that I think are on par with Eggert's CS 33. It is very hard to follow sometimes and I have no idea what's important to write down. Basically, If you don't spend time reading that textbook, you will probably have trouble understanding concepts.

The first two homework are too easy (Lisp coding that can be finished in 10 minutes), and the last two are a disaster.
Homework 3 as far as I know no one received 100. Basically after spending several hours optimizing routines and heuristics will give you a score of 94.2 instead of 94.
Homework 5 and 6 are like math questions. They are pretty good, and they really help me understand what resolution, markov assumption, etc. are. However, solutions are never posted, and we pretty much have no idea what's going on after we submit our homework. Due to students' constant requests, the TA inputted what questions we did wrong and how much points were taken off like near 12am of the final exam. But, I can't comment on how helpful they turn out to be.

The final exam is all multiple choices. It is very anti-south campus. If you are good at memorizing stuff, you can get 90+ for not understanding any math, logic or coding; If you are not good at it, like me, you are most likely screwed.

In the end, I am pretty frustrated with this course. This AI course seems too traditional. It focuses too much on search (why not talk about it in 180?) and logic (not everyone has taken Philosophy 31), and talks too little about current learning algorithms. Homework and exams are probably reused so solutions are never posted, which means if you did something wrong, you will probably be still confused about why you did it wrong after the class is over. In the end, I can safely say I hardly learned anything after struggling through the whole quarter. As a reference, I have a 4.0 GPA for all CS classes before this quarter.

Originally posted a positive official course evaluation, changed my mind after the final.

So many things pointed to it being a practical problem solving exam, including this quote from the study guide: "Questions will test for insight and understanding. I will not ask historical or encyclopedic information", the midterm and last two homeworks (and associated lectures) being practical problems, and the allowance of calculators on the exam.

Guess what the exam was on.

Yes, he's not obliged to make things easy, and all the questions were things that were taught, and it probably wasn't intentionally malicious, but if you take his class beware surprises or trying to guess what might be on the final.

Rant aside, he's an alright lecturer, teaches from the book but keeps things logically connected (in the sense that he teaches the book's topics and examples rather than reading paragraph summaries). If you don't have the book, you can do alright off of his lectures, but his examples might have no lead-up since they're from the book which spends a bit more time laying out the problem.

Homeworks are in lisp, except the last ones which are practical problems. Overall they're pretty great in that they give an intuitive feel for the topics (besides final, see above). Generally, they have good "pacing" in that you can finish one function at a time, and see how they all come together. Minor quibbles with specific homeworks, one had a heuristic competition aspect, where only #1 got 100%, and the slowest got at most 80%, another had no way of telling if you were on the right track since you plugged outputs into a blackbox solver. The problem sets were good questions, but no solutions were posted afterwards, making them a bit useless for studying.

Class materials were not too great, the study guides had some issues besides the above, they were basically the syllabus list with a little more detail. No slides or lecture notes, the study guides mostly just let you google AI topics without getting bogged down in sci-fi or news articles.