#198 overall
CS221: Artificial Intelligence: Principles and Techniques
Stanford University
2 recommenders · 2 threads · recommended in 2 years, 2022–2023 · score 2.2
Stanford's introductory artificial intelligence course, taught by Percy Liang and Dorsa Sadigh in recent offerings, covering core AI techniques such as search, probabilistic models, games, and machine learning fundamentals. Commenters point to lecture recordings and a free online version.
Commenters call it practical and kept up to date, with strong lecturers, and say it is a required companion to CS 229 for Stanford's AI track.
Machine Learning & AIintermediatefreevideo lectures#artificial intelligence#search#machine learning#percy liang#dorsa sadigh
Recommendations by year
20062026
Gilbert Strang's lectures on Linear Algebra [1]. Prof Strang is an amazing lecturer with a unassuming style. He's the expert on Linear Algebra and makes the topic so much approachable.
Ken Joy's lectures on Computer Graphics [2]. Prof Joy is another amazing lecturer, making the Computer Graphics topics seeming easy.
Stanford CS221 Learn AI (2019) by Percy Liang and Dorsa Sadigh. Both professors are great lecturers. Andrew Ng's ML class was great, but it was more academically tuned. CS221 is more on the practical side and is more updated as ML is progressing fast.
Micromouse 2021-2022 by UCLA [4]. It's a short series taught by graduate students and probably it's incomplete, but the content and teaching are amazing. I wish I had this kind of class when I was in school. The teaching and materials are very approachable and easy to understand. It shows how basic electronic components and basic circuitry work. It shows how to put them together and how to write simple programs to control the components. The end result is a robotic mouse that can traverse mazes with seemingly intelligence.
[1] https://www.youtube.com/playlist?list=PL49CF3715CB9EF31D
[2] https://www.youtube.com/watch?v=01YSK5gIEYQ&list=PL_w_qWAQZt...
[3] https://www.youtube.com/watch?v=J8Eh7RqggsU&list=PLoROMvodv4...
[4] https://www.youtube.com/playlist?list=PLAWsHzw_h0iiz1EQEvQ9n...
Edit: added [4]
Stanford grad courses are excellent. Just following the exercises will get you very far:
* AI Foundations: https://stanford-cs221.github.io/autumn2022/
* Meta Learning: https://cs330.stanford.edu/
* Theoretical ML: https://cs229.stanford.edu/
* NLP: https://web.stanford.edu/class/cs224n/
* Computer vision: http://cs231n.stanford.edu/
* Reinforcement Learning: https://web.stanford.edu/class/cs234/
Read Blogs:
* DeepMind: https://www.deepmind.com/blog
* Google Research: https://ai.googleblog.com/
Replicate papers that you are interested in with a twist specific to your use cases
* RL Spinning UP:https://spinningup.openai.com/en/latest/
Read “AI a modern approach“ and do the MIT and Stanford courses that are available in Youtube. Then you can go deeper into the branches presented in the book and courses. The problem I’m seeing now, ist that everybody seems to think AI/ML is NN. Nothing further from truth!
Stanford CS majors who take the AI track are required to take both CS 229 and this course, CS 221, so it definitely presents useful materials and concepts.
I want to address the following quote from the article: "During the fall term, the Stanford students taking CS221 preferred watching the KnowLabs videos. Thrun says this improved their performance."
I took the official Stanford offering this past fall. I am not a Stanford student (anymore), but I enrolled as part of SCPD. As a result of the official course starting two weeks earlier than the public offering, the course staff usually didn't post up the YouTube videos in time for me to do the assignments, so I relied primarily on the lectures. (Actually, as a distance student, I relied upon the lecture recordings.)
I eventually did catch up on the YouTube videos, and I can attest that they are indeed better. However, they produced superior results mainly because they covered more material. For example, in one case, I did over two pages of error-prone calculations because I didn't know the mathematical shortcut that was taught in the YouTube videos. And I gasped as the professors spelled out in detail how to do problems that I had to think hard about how to do. Indeed, one problem that I was pleasantly challenged by on the midterm was simply not covered in lecture; in contrast, the professors explained it tutorial-style on YouTube.
Using the terminology of the course, I would say that the YouTube videos "overfitted" to the homework and exam problems -- so of course the students preferred the videos!
I do suspect that short, interactive, and extremely polished videos will be proven to be better in most cases, but unfortunately this experiment cannot be used as a basis for forming that conclusion. Clearly, the lectures were handicapped.
And to be honest, I am skeptical of how videos would perform against some of the tremendously excellent lectures out there, such as the introductory computer science lectures that are posted at Stanford Engineering Everywhere.
With all this said, I had a great deal of fun taking this course. I applaud the course staff for doing a stellar job, and I am grateful to be part of what I am certain will be a historical experiment.