Tuesday, July 19, 2016

National Academy to reform science in US?

Some folks I know, rightly proud of their Administration connections, recently mentioned a new project at National Academy of Sciences aimed at a particular new reform of NSF (and really of science in America). My (quick personal) response to them:

 Good morning, ...!

For a few weeks, almost all the straws in the wind have been pushing me
more strongly to support Clinton over Trump. Your message is the strongest impulse I have received so far in the opposite direction. 

More precisely, after having worked for NSF for more than 35 years (until retirement last year), I can't help reading between the lines in what you said [about this project]

What kinds of organizational structures lead to maximum productivity and development of human resources in the S&T sector? There have been political speeches about that for many many decades, but so far as I know the book Leadership in S&T from Sage edited by Bill Bainbridge (who is probably still at NSF) is one of the most definitive. I see it as very compatible with another book from Sage by Ashkanasy et al, on organizational culture and productivity in general. 

"Reform" has become quite a slogan for the past recent years. There was such great support for Carly Fiorina based on the kind of reforms she carried out to Bell Labs, one of the topics covered in Bill's book. After Bells Labs and many of its siblings were truncated out, continued US leadership depended very heavily on a system of free universities, NSF, DARPA, and the "XROs"

And then came the great reform of Bush and Cheney, under Tony Tether, which really began to bite in 2003. I remember 2003 very well from personal experience. The great reform under Tony Tether was accompanied by many great speeches and powerpoints on "how to transition new technology" which I won't elaborate on here, but which certainly remind me of new things 
done to NSF.

IN 2003, I was one of the eight or so program directors in the electrical engineering division of NSF, the division responsible for fundamentally new breakthrough research in areas including electronics, photonics, communications, computational intelligence, and electric power grids (which were funded by my division long before the Office of Electricity Delivery of DOE  put big money into a different approach). For years, I had been able, at the end of the day, to fund about 30% of the proposals coming into my area, proposals which already represented much more advanced than average creative new ideas from universities across the US. (Small businesses were also an important part of the story, but this email will be long enough as is...) I remember very vividly the "Halloween massacre of 2003", when it dropped very abruptly to 10% and when a new division director announced that the cut would be accompanied by  her micromanaging all funding in order to help reduce and filter what was funded. 

The transition was slower in other parts of Engineering and of Computer Science,  but just as severe. Many of us worried: "If it takes huge effort to put together a worthwhile proposal, and your chances are only 10% of being funded, won't people just drop out, and go back to an equilibrium with a higher rate but a lot less research?" It was trending that way until the stimulus bill in 2009 (which raised hope and proposal counts), but I think that was a blip. When I left, it was still no more than 12% for major, competitive proposals (asking typically for $100,000 per year over 3 years). 

LOTS of serious intellectual effort went into analyzing why there was such a huge drop in success rates for core, competitive proposals. (By the way, statistics for noncore proposals, like 5K travel grants or undergraduate student supplements, PARTIALLY masked the data on the drop in the  NUMBER of proposals funded. It was every bit as bad as I described.)  
But everyone agreed that one major factor was something very obvious to us from day one: we were getting a whole more proposals from "refugees" from NASA and DOD. Whereas before, NSF handled about 30% of the funding for core innovative EE research in universities (the foundation of those great US university programs famous around the world back then), and DOD+NASA handled most of the rest, the Great Reform at DARPA and Shelby's similar interventions at NASA had created a stream of "refugees," of people coming to NSF for funding. Crudely: if people needing you are suddenly tripled, it should not surprise you if proposals triple.

Back then, Tether argued that we did not really need DARPA for the basic research part. There were "others like NSF" who could handle that. And so, making certain corporate stakeholders very happy, they decided DARPA would be more for translational technology for them. The universities protested vigorously, and Tether replied that universities could still get SOME funding -- basically as subcontractors to those big stakeholders, using them for well-defined specific pieces of work in more vetted, proven, translational areas. And thus we had a refugee problem. It was even worse at NASA.

A key point here is this: translational research can be great if there is something to translate. If not, it is only a matter of time when the pipeline dries up. I found it especially poignant a few weeks ago, when JOhn Mankins
(former head of HR&T, NASA's main new technology thrust under Steidle) said approximately: "It looks as if we CAN get launch costs low enough to meet the requirements for 9-cents-per-kwh electricity from space, thanks to the advanced launch technology now being developed in India..." 
(By the way, if you think solar is already at 4 cents per kwh, I posted a link at LinkedIn to a video of the president of the World Bank. Or you can google on the numbers in EEG.) India? There once was a time when the breakthroughs came here. But yes the US leads in commercials and dancing girls for our new space companies. (Not bad dancing girls!!) And innovative ways with manipulating numbers are part of our newfound superiority in buying other people's products. There are amazing similarities between some of the enthusiastic new translational efforts and enthusiatic things that Bo Xilai was doing in China, but people have been warned that these are national secrets because of the great competitive new edge they give us.

And so, another nail will be hammered into the coffin.

Maybe it is just as well. Some of the new areas of technology we were opening up could get to be very dangerous, as in the Terminator movie or worse.   (The new wave of deep learning translation was mainly really due to
Andrew Ng, Y. Dan, E. Boyden and Y, LeCun, EFRI-COPN Deep Learning in the Mammalian Visual Cortex, http://www.nsf.gov/awardsearch/showAward?AWD_ID=0835878&HistoricalAwards=false
which was just the first step of many which might have come. Still, China and the EU have continued activity in many such areas of technology, and there may yet be surprises coming. 

It is ironic to me personally that technology translation really was a major concern of mine, back before the reforms started to bite hard. But I don't see how funding for dancing girls and traditional beancounting would solve the many problems I saw in those areas. I see no alternative to the kind of very hard work some of us would try to do in the panel review system, which I once compared in my mind to auto engines which were only 25% efficient but were the only things actually able to move back then (though later our work was changing the picture on engines too, until oil stakeholders objected).
I also compared that good old competitive system, in the spirit of Vannevar Bush, to what Winston Churchill said about democracy: "The worst system anyone has ever devised, except for everything else anyone has ever tried or devised." As ugly as making sausage at times utterly dependent on human beings and human cultures, fallible, in need of intelligent improvement ..
even more competitive than the most extreme private fund competitions but more open to the high-risk cutting edge... and now...

well, the reforms remind me a lot of what Schumpeter predicted, the prediction that society simply would not put up with protracted competition.
I often think about the movie The Aviator (about Howard Hughes) and about the folks who almost shut down Edison when he was about to turn on the lights in New York City... and how the predictions of Schumpeter and Spengler do seem to be going by the numbers. But for folks who are used to a competitive IT private sector... don't imagine that the new reforms have nothing being prepared for you. 

Given such changes, I found it a bit challenging when IEEE asked me to give the keynote talk to their winter summer school on computational intelligence in big data  a few months ago (after my retirement, based on my personal work in that field). At www.werbos.com/Mind.htm, I posted the slides, which may be more digestible and quick than the long words of this email. 

By the way, ALL jobs can be automated. For folks who think a certain way, there is no need for any human beings at all, in space or in earth. That way of thinking is a very serious problem, in my view, exactly as one of you recently suggested, starting from 1980 as a transition year. More precisely -- the combination of new technology and a will to misuse it is like the combination of an automatic rifle and a particular metric of productivity in using it (as in Texas recently).  


[long as this was, one could write a whole book to fill in details... but not by me this morning!]

No comments:

Post a Comment