Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

Up until around five years ago, the course 6.034 had a Fall version and a Spring version. They were substantially different. In the Fall it was taught by Patrick Winston, and in the Spring it was taught by other professors who were 10-20 years younger than Winston.

It's my understanding that the Spring professors really disagreed with the content in the Fall version. It was too old-school AI and they believed it was a disservice to students to teach them AI in that way. I believe they marginalized Winston.

To be honest, I cannot tell if they are right or wrong. I am not familiar with the curriculum in Winston's 6.034. How much does it look like an AI textbook from the 90s? I know that he included topics such as SVM, which is, relative to Winston, newer AI.

But, in any case, I do wish that the two philosophies didn't butt heads. Especially because I have no doubt that Winston's course was better organized, and had more enjoyable lectures.



I think this sentiment certainly existed among some of the younger CSAIL professors- in fact I interviewed for a UROP once with an unnamed professor and they asked "what related classes have you taken so far?" And as soon as I said 6.034 (before I could even finish the list) they said "I'm sure that class is fun but so is a cooking class, it isn't relevant here".

But I'd like to think it wasn't the prevailing sentiment in the department. I actually think a little bit of it stemmed from jealousy about the student praise PHW received. Because I know plenty of students that are now doing more "modern" or "hardcore" ML research that praise 6.034 highly.

It's true that 6.034 did not teach enough math to jump right into theory research, and it's true it didn't teach enough modern implementation to go right into an industry application. But it certainly didn't look like a textbook from the 90s! Professor Winston reworked the course a little bit every year. He gave excellent lectures on the concepts behind deep learning and reinforcement learning. He brought in current professors to present their research goals almost every Friday.

It covered plenty of modern topics, just in a more abstract sense. Which is perfect for an intro course, especially one that is taken by not just CS students but those from many departments (even graduate students in architecture, business, etc.). I think it's so helpful to start with understanding these concepts and then work through the details, rather than start in the weeds like many other courses do.

He did also spend time covering older concepts, which are still good to know, especially when presented through the lense of the history of the field (in academia and business). That's just not something available in any other courses that I'm aware of. Same thing with his amazing communication advice and his sharing of interesting biology, psychology, and neuroscience research.

I sincerely hope 6.034 will continue on somehow in the spirit of the original course. It was inspiring to many students, with life/career lessons that are memorable well beyond school. MIT has plenty of good more traditional ML courses, as do many schools. What I haven't heard of elsewhere is something like 6.034. It was one thing that made MIT unique I think.

Anyway, that is rambly, but I just wanted to defend not only the execution, but the philosophy of 034 (or my take on it anyway, can't speak for PHW). And I hope that the minority that dislikes it doesn't try to take over the course now.




Consider applying for YC's Fall 2026 batch! Applications are open till July 27.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: