Startups don't do research. Some may try to commercialize research. Some research labs may also try to commercialize research. But true research is not something anyone would risk VC on. True research takes decades to see any ROI and usually even then it is unclear exactly what the ROI is, making traditional funding metrics useless. Here is an interesting book about Dupont, IBM, and other research labs trying to quantify the output of their research labs:

http://books.nap.edu/openbook.php?record_id=6200&page=6

Bottom line: you have to invest in research because you know it will pay off. But it is impossible to say exactly how or when it will pay off. The only kind of institution that can support that kind of investment is government or large (probably monopolistic) corporations.

Not that the actual research part has paid off, but companies like Anybots have been much more successful in their bipedal research efforts (see Dexter @ http://anybots.com/abouttherobots.html) than many academic labs, much less other companies.

Anybots is probably not a traditional company in terms of the desire to make money ("maximize shareholder value").

I think this concept of a traditional company doesn't really exist, or is fairly rare.

Most companies care about their employees, and care about their work. They aren't trying to get every $ they can.

Correct. Starting a company is fun, but it's not a time in which you can do research. Research is inherently risky and long-view oriented. Because startups are inherently risky and must survive in the short-term, the added risk of structuring a company around a research project is likely to lead to so much risk that no one will want to fund it.

Well-funded government labs can do R&D because they're in a low-risk position and can take the long view. A government researcher is going to keep his job regardless of whether his findings contribute to the immediate P&L, so he's able to pursue research projects that are unlikely to pay off in the short term.

Contrary to the common slight against government regarding innovation, government research projects can be highly innovative and useful. As for CEOs claiming that R&D departments are less innovative than private-sector executives, well... they're CEOs, so of course they're going to be immensely biased in favor of their own kind. Their definition of innovation is based on what they recognize as innovation, so of course most of them will underappreciate the long-term-oriented contributions of R&D.

The problem with academia, though, is that it has the worst of both worlds. It's bureaucratic and stuffy, moreso than most government research labs, but the level of risk (career death) for the aspiring professor is intolerably high, meaning that high-risk research projects are unpalatable. Academia's a sinking ship, and no longer a viable career for more than a handful of people per year. I hate to say it, but academia's best days are almost three decades behind it, by this point.

His timeline doesn't apply to Computer Science - and possibly other fields like engineering, but I have much less experience with them. Because CS Ph.D.s can get industry jobs, there's less need for them to do postdocs. Some do, but it's still common for freshly minted Ph.D.s to get academic jobs. Of course, getting those positions is still just as difficult as it is for the real sciences, but there's not necessarily the 2-6 year delay.

I've given this topic some thought recently. Academia has it's problems, but what I do like is that I can work on things that are interesting and useful, but not necessarily profitable.

How can something be useful without being profitable? I do research in systems and high performance. In theory, the software we create could be sold. Some startups exist that do just that - Cilk Arts and RapidMind are two examples. But I'm skeptical of their viability; companies that exists solely to sell compilers rarely do well.

The kind of research I do is valuable to companies, but rarely as something to sell on its own. Intel, for example, does research in programming models for emerging architectures not because they want to sell that software, but because they need it so programmers can use the processors they do sell.

I'm suspicious that the measure of "innovation" is the number of patents filed, but I can't confirm that from the article. In academia, good ideas rarely materialize into patents; they become papers, and the relevance/influence/prestige is quantified in the number of other papers that cite them.

Fortunately, many schools do support startups (financially, even) coming out of their grad schools. I haven't noticed any discussion on the success of this kind of support versus traditional VC, but it makes lots of sense for a grad student to turn a thesis project into a startup immediately after graduation.

There is a lot more happening on the enterprise/university front. My new startup took pre-commercial funding from a fund set up between a consortium of universities. I think this is a potential solution to the problems the blog author identified, but for the funds to work properly, they need to be administered and run by business people, not academics, and especially not academic administrators.

Asking CEOs about sources of innovation seems rather biased. I think it is rather difficult to tackle difficult problems of certain kinds in a startup. Usually, you want to have an idea before you start that you can succeed.

Take the face recognition thing that Apple just announced. OK, that one came from a corporation, but one with lots of money. I think for a startup it would have been difficult to say "let's create a startup about face recognition in photos" because until recently, no sufficiently good method for doing it was known.

They should add competitors - lots of companies are simply first followers.