Monday, February 23, 2009

On the best use of stimulus research dollars

John Langford has a post up advocating that research dollars ought to go to industrial research labs. My first temptation was to write a counter-post advocating that research dollars ought to go to university researchers in less-than-well-known mountainous locations, especially pre-tenure refugees from industrial research labs hungry for the money and willing to do lots of work.

But more seriously, I think there's plenty wrong with John's argument. He argues that the Bell Labs model of the 1900s and earlier has the main components of a successful research agenda: access to cutting edge problems, free time for researchers, and concentration, and infers from this that the best use of government funding is to fund basic research at companies that have such labs (MS, ATT, Lucent, IBM, Yahoo, Google are cited).

Along the way, he throws out various statements that I have problems with:

Some research universities manage to achieve at least access and concentration to some extent, but hidden difficulties exist. For example, professors often don't work with other professors, because they are both too busy with students and they must make a case for tenure based on work which is unambiguously their own.

This is partly true: there's less collaboration in universities than in labs, but much of the collaboration in research labs is also by necessity: to get some things done takes a number of people, and you don't really have access to a ready supply of slaves students. Collaborative research is not by definition better.

at least research at national labs have had relatively little impact on newer fields such as computer science.

National labs play an important role on lots of large-scale visualization work (I know this because I'm at one of tthe best viz places in the country, and they have extensive collaborations with national labs). National labs compete like universities for research money, and often have the inside track on funding from places like the DoE. Their sweet spot is the kind of large-scale infrastructure work that's hard to do at universities or at industrial labs.

Some people might think that basic research done at a university is inherently more desirable than the same in industry. I don't see any reason for this. For example, it seems that patentable research is about as likely to be patented at a university as elsewhere, and hence equally restricted for public use over the duration of a patent. Other people might think that basic research only really happens at universities or national labs, but that simply doesn't agree with history.

This is a strawman: I don't know who 'some people' is, but I think any reasonable position would argue that the kind of research is different: someone once told me that the ideal industrial project involves 3-5 people: any larger or smaller is best done at a university. Whether the research is basic or not depends on the work: I don't know of many industrial labs that support research in complexity theory (except MSR) which is arguably basic research, but there's very fundamental research done at many labs in areas like auctions, ad modelling, large-scale computing, and so on.

But quibbling over statements aside, I think that the best use for stimulus money is neither universities (though I'm very grateful for the $3B) or industrial labs. I think the best use is in places where its already going: the so-called "shovel-ready" projects in green tech/renewable energy. The kind of funding that leads to direct economic impact is not going to come out of either universities (which naturally take a longer time line) or industrial research labs (that have lots of sloth and bureaucracy). It's going to come from VC-type funding for energetic startups that actually make things happen. Yes, there'll be research, but presumably the projects being funded will be well beyond the research stage and ready to make things happen now, or in the next few years. That is after all the point of the stimulus.

The budget will be coming out soon, and I hope that attention is paid to a longer-term reversal of the depredations in science funding in the US. But the stimulus package is about the present and the short term.

Other notes:
* In the comments, Hal argues that education is an important mandate of the NSF, and that's why resources get channelled to universities. I'd also add that I don't see why taxpayers should fund a corporation's bottom-line: money for Yahoo helps Yahoo, whereas money to fund students helps generates more expertise.

John says: "In economic terms, these companies have for reasons of their own decided to provide a public good. As long as we are interested, as a nation, or as a civilization, in subsidizing this public good, it is desirable to do this as efficiently as possible.". Permit me to snigger. Bell Labs had a monopoly on the telephone network for eons, and having a research arm was good PR for them. Once the monopoly collapsed, so did the dedication to provide a public good. Having worked at AT&T these many years, I am deeply grateful for the opportunities I had there, but there was a clear focus on research that helps the company bottom line. Even the much vaunted Google Research makes no secret of its focus on company-specific projects (the 20% rule implies an 80%!) (disclaimer: I occasionally consult for Google).

Post your comments here, or over at the original thread.


  1. I don't know of many industrial labs that support research in complexity theory (except MSR)

    Well, IBM Almaden was once famous for its lower bounds focus (you're not going to tell me Ajtai is there for the applied database research).

    One can argue that Peter Shor's presence at ATT also had more to do with basic complexity theory than the company.

  2. Ouch ! now that's an embarrassing omission. In my defense I guess I was thinking of a more 'structural' support for complexity, but even by that account Almaden had/has a good crowd.

  3. One could argue that David Johnson, Mike Garey, and Mihalis Yannakakis did a fair amount of complexity theory at Bell Labs. Madhu Sudan was at IBM Watson for several years.

  4. "the 20% rule implies an 80%!"

    I think you are confusing things. The 20% rule has nothing to do with research, so far as I know.


Disqus for The Geomblog