It's the course that launched Coursera (formerly ml-class.org), and is still one of the most highly rated on Coursera, so I dare say that you are in the minority with that opinion.

Care to explain why? Anyone else who did it care to chime in?

I took it several years back and enjoyed it. I liked that the course had you implement the whole training pipeline yourself rather than using a framework (not sure if the newer class does the same). While you would likely not do this in practice, I felt it helped my intuition when using the frameworks since I had a sense of how the internals were working.

It was good like 10 years ago and it did age well but it's a little bit outdated on the video/audio quality and the tools and algorithms you learn about. I think it's surprisingly up to date for a fundamentals course that old, but still a bit outdated.

I'm doing it along with Ng's newer courses at the moment and I really like that he focuses on all the basics mathematically as well and not only conceptually which gives you a deeper understanding machine learning imo. However as others have said, the audio quality is subpar and personally I find it hard to motivate myself for the programming challenges in Octave. So my suggestion would be to just view the videos and take notes and then do the newer courses and their challenges.

Due to the low quality of the video and audio I honestly struggled to want to go through the material.

I second that - great content, but terrible audio / video quality, unfortunately, which makes it less enjoyable and a bit of a struggle tbh.

It requires Matlab for instance.

I completed the course with Octave, but yeah this language is a hurdle that people don't need.

back then, matlab was the thing

actually, Matlab is still the thing depending on the domain you are working with. I don't get the hate towards Matlab generally from CS people. Maybe because it's paid?

Rather it is because it is a poorly suited language, in that isn't aware of modern programming approaches.