> There are plenty of ways to bring smart people together to learn and do research.
  > They could be nonprofit; they could be separate from teaching or integrated
  > with it. You could even call them "universities" if you like. But there's no
  > reason they have to resemble the current monster.

Can you name some ways to actually do this? I'd love to hear about them, but have never actually seen a system like that. Even if there was though, before you go yanking money out of the system, you should actually try to replace it. And that will take time and money, and you will have to convince great scientists to make the switch, in spite of the fact that they will probably have less money, less resources, and spend more time doing something other than interesting research in order to build said system. That's all really hard.

Recently a lot of academics have been pushing for open journals, even going so far as to boycott closed ones. Even this is fairly controversial. :(

  > Sure, but the university system isn't sold that way. We're meant to believe
  > the research done in universities is all useful.

Sold by whom? Guidance counselors? Other students? I don't believe I've ever met a professor who thought that all, or even most research, is life-changing, or even very interesting. I just don't buy that you heard that from a reliable source.

  > Universities have special privileges when it comes to fostering immigration.
  > If America had world-class research centers with the same privileges,
  > people would still love to come here and they'd be able to do so.

It's not only about privilege: schools are known by foreigners as a system that allows them to get to the US. In order for this transition to be effective, you'd have to supplant the educational system, then get everyone to know about your alternative system.

And besides that, the institutions that do the hiring are incredibly biased towards American schools. I work for a reasonably prestigious lab, and the people in charge are (charitably put) deeply suspicious of research from Chinese and Indian schools. When given a choice, they will hire a PhD from an American school almost always. Exceptions include University College at London, Oxford, Cambridge, Utrecht, etc., but it is a vanishingly small list, and even they I would say are much less likely to be hired.

  > As you point out, universities aren't very good at this. Maybe a lot of people
  > won't use libraries, and many who do won't be very productive. But
  > as you suggested in one of your other points, the instances that pay
  > off really pay off. And libraries do this much more efficiently than universities, in terms of money spent. There are a lot of
  > university students burning tens of thousands of dollars without learning
  > very much.

Oh man, I'd really love to believe that, but I just don't see it. Science is still pretty much a communal affair, and the mere availability of knowledge does not make science happen. You need a robust social framework to evolve it. I just don't see that in libraries.