Winter 2019 - I can see where the negative reviews come from regarding the course content since ppl would expect AI courses to be modern and fun instead of theories. While I agree with that, I do want to add my personal thoughts regarding the problem. There are also other professors besides prof. Gu who teach this course and cs department requires them to teach the same materials (otherwise it would be unfair for both teachers and students in different quarters). this intro level AI course was designed years ago and ofc it is a little outdated, but the content can hardly be changed unless the department decides to. I guess for ppl complaining here, it would be better if you talk to cs dept directly instead of giving a low rating for some professors... Regarding the professor, I took the course when the pandemic hit in 2020 and everything was a mess. I think the professor is knowledgeable and cared a lot about course quality and did a great job accommodating students' needs. I do agree that sometimes the slides are too brief and the textbook definitely gives a more thorough explanation. BUT that is based on if you don't listen to the lecture at all and just reading the slides. Based on my personal experience, it is easier to understand the materials when I went to the lecture with professor's demo. For TAs, I would agree that they were not that helpful comparing with TAs from other courses. but I do not think they were being lazy (at least mine wasn't) Their speaking skills are not too good so it's difficult to understand, but they were willing to stay after discussion with me to make sure my concerns were resolved. The HWs and tests are doable as other comments said. In general, I think this course is a descent intro-level AI course that shows/prepares you the fundamentals behind the fancy side of AI/ML. I also consider it as a good elective with very manageable workload and easy A.
Fall 2022 - Lectures: He is not the best lecturer. For the more mathematical/theoretical content, I learned more from the book. For the content on neural networks, he didn’t even seem familiar with the slides. But the networks content also is only on quizzes, so I guess it’s ok. Homework 35%: They are very mathematical questions until the switch into talking about networks, then the homework becomes implementing networks by scratch in Python. It was also a grind for the math homeworks because they were required to be in LaTeX. Midterm 30%: They generously gave a very similar practice midterm, so as long as you can confidently solve those problems, or at least note how to solve them on your cheatsheet, you’ll score well. It also covers the math content more. No final, 6 quizzes that are only worth 5% total of grade. Final project 30%: You can choose from some given proposals that Prof. asked people he knows to provide some ideas for, or pursue your own project if you’re smart. Would recommend trying to take the course with friends/colleagues you trust, so that you can collaborate with no issues. Overall, I would recommend this class if you are mathematically mature, have an interest in the math in ML theory, or are just very ML-invested. Or have an interest in self-learning all the math and network concepts. I unfortunately have little qualification in any of the above, so I suffered silently. Sunk cost fallacy is real. Shoutout to Lucas Tecot for the HW hints.