The consensus here seems to be that, unless you want to do research and publish papers, a Bachelors in CS is a terminal degree. Is there really no room for formal exposure to advanced application of what the researchers are researching?
If I think machine learning sounds like an interesting approach to solving problems, are my choices really limited either to seven years of academic research or to teaching myself this stuff from whatever materials I can cobble together in my spare time?
The point is that practical experience in a Master's that offers research work is more valuable than a coursework-only Master's. Graduate courses are usually easier to complete than undergraduate courses, and they are excellent for complementing a research program.
If you want to do machine learning, go and do a Master's with a thesis option where your advisor does research in machine learning project and where you will be encouraged to publish your work at conferences.
To my mind research is fundamentally different from (and antecedent to) application. I assume (perhaps mistakenly) that there are areas of knowledge that are both potentially useful to solve problems and difficult to access/learn on ones own. Why is it not reasonable for that advanced knowledge to be collected, distilled, and taught?
Further, from my brief exposure to a CS grad lab, research occurs in very narrow areas, often on things with no discernable utility, where all work is essentially throw-away, and where the only goal is to publish a paper. In contrast, I want to build stuff, possibly hard stuff, stuff that someone would use; and that may mean learning things (not discovering things).
From the perspective of industry, practical experience building research toys is generally more valuable than either book learning or course projects, because you are forced to come up with and implement novel solutions and because you ultimately have to make something work in order to get the degree. This is the value of a Master's with a practical thesis.
It's not that the advanced knowledge isn't taught in a graduate program, it's that a formal research project where some or all of the advanced knowledge is applied is seen as a great way to integrate everything, and it's completion is a better marker of autonomy than course project completion. Of course there are professors doing useless research, so find someone at a school doing something that interests you. Or get hired by a company and learn on the job. Or start a side project. Yes, academia can be a waste of time if you don't actually like the work, but depending on what you want to build it might be the easiest way to get there. Grad school is generally an opportunity to become an expert in the field that interests you, and the production pressures of industry might preclude this.
In my experience, employers care more about research than you might imagine (for reasons the other replier has mentioned).
That said, a coursework MS will make you generally much more attractive to employers. It's just that they will interpret the degree as more of a signal of your capability and intelligence (and this effect is greatly underestimated by HN) than an indication of any specific expertise.
So with a coursework masters, you stand a chance of having generally great job opportunities, but you may be disappointed that you are not eligible for the most interesting jobs, even though they are related to your coursework.
If I think machine learning sounds like an interesting approach to solving problems, are my choices really limited either to seven years of academic research or to teaching myself this stuff from whatever materials I can cobble together in my spare time?