There are tens of thousands of students taking these courses. There's probably a business opportunity here. How could a startup make money by hiring, say, one hundred of the best Machine Learning graduates?
the Machine Learning class (at least for the Stanford students) also has a project component, so I'm sure a startup could see some use even from just having the graduates work on their problems for a research project, if nothing else! personally, I ended up doing two projects last year with data from my previous internship, and I think it's a great way to both have interesting problems and useful results.
I'm somewhat sad that the ML course project isn't an option available to the online students. I know there's no way something like that could scale, but I'd still like to have more validation that I know what I'm talking about than 'yes, you can answer ?'s and do programming assignments.'
When he found out I'm taking this course (he's a Stanford alum from the 90's), the CEO of ThisOrThat, the startup I work for, was really excited. He said I could use production data for a project if necessary, and gave me some ideas for possible ML applications that would be really helpful to have on the site (fraud detection, a reputation system, etc). I'm looking forward to this more than anything I did in college or grad school, and I actually feel like I'm on the young side of 25 again.
Even so, I'd really like having someone who groks ML giving me pointers after the fact. 3 months just doesn't seem like enough time to do a thorough dive into the material for me to say with confidence "I understand the background and problems in the field of ML".
I definitely agree with your last point - part of the reason why I decided to take a machine learning class again from Prof. Ng was because I still had a sense of unease that I wasn't quite getting the topics that would really be important when I'd implement machine learning algorithms (whatever they end up being).
Mostly though, I'm just hoping I can find a place to work with plenty of ML people to learn from :)