FEATURE ARTICLE
Science in 2006, Revisited
From grid computing to genomics, the science fiction of 1986 is fast becoming science fact. There remains equal reward in the signal and in the noise
Lewis Branscomb
Clear Crystal Balls
In a few cases I made attempts to be so extravagant that my forecast
would be seen as tongue-in-cheek—and therefore turned out to
be about right for 2003. I foresaw teraflop computing (trillions of
calculations per second) available on the desktop. The development
of new distributed operating systems that allow large numbers of
computers to share distributed data (called grid computing) and to
share the computing power of machines that are momentarily idle
joined with Moore's Law (that processors double in speed every three
years) to make this possible. I correctly forecast the mapping of
the human genome in the 1990s, and predicted that you would be able
to buy it on a CD-ROM for $9.99. Most important, I predicted the
growth of the Internet and its impact on science. (In 1986 Steve
Wolff came to the National Science Foundation and launched NSFnet
using TCP/IP, the key protocols that permitted the explosive growth
of the Internet later. Looking back, it is hard to believe that it
was not until 1987 that 10,000 Internet hosts existed; now there are
hundreds of millions around the world. See timeline at
http://www.pbs.org/internet/timeline/) I called it WUNET (an acronym
for World University Network and pun on "one net"), making
a serious underprediction of the commercialization of the Internet.
On one point I was actually too pessimistic (although literally
accurate) in predicting that by 2006 automatic language translation
would "remain incompletely solved." Finally, I correctly
predicted the confusion that would engulf tenure and promotion
committees in their attempt to define publication so they could
decide who should perish. Even in 1986 it was clear that authors
would become publishers, and scientists would not wait to learn the
latest research advances until the print materials arrived in the
snail mail.
But these were details—some of them intended to be funny when
looked back upon. The serious part of the prediction about science
itself was in one way correct and in another unrealized. All the
trends for science were evident in 1986, and I believe most
scientists would endorse the observations I made. But most would
also say that the majority of science and its institutions have not
moved from their traditional self views. Change happens blindingly
fast in science, but agonizingly slowly in the institutions
of science. Let me summarize the most important trends I forecast.
First, science would become ever more
capital-intensive, which itself would drive science
down a multidisciplinary, multi-author, shared-resource path. That
was a no-brainer; it surely is reality. I did propose a solution to
the competition among nations for the location of the very large,
shared, science facilities. Every participating nation would be
authorized to build a magnificent marble structure, with
"World's Largest Accelerator" or "World's Most
Farsighted Telescope" engraved over the front door. Inside,
there would be of course no accelerator or telescope, only a mammoth
bank of computers and satellite dishes through which each nation's
scientists operated a machine in a generally unknown place, deep
underground or atop an inaccessible mountain. Site selection becomes
much easier. This trend is also well under way.
Second, I saw the reintegration of
sciences —a hugely important trend that is
surely under way but will be far from dominant in the structure of
scientific activities in 2006.
This trend can be seen in at least four areas:
—Cognitive science, brain studies, neurophysiology and
behavioral science. In these areas we do see a huge effort to bring
together these several threads of knowledge, much of it based on
faith that surely one day we can give biochemical and physiological
understanding to human (and animal) behavior.
—Cosmology, high-energy physics, astrophysics and
mathematics. Here too the prediction is in full flower. Indeed,
except for mathematics these disciplines now find themselves in the
same department in many universities. Progress has been nothing less
than incredible. And the prediction that the estrangement of
mathematics from theoretical physics would end has surely proved right.
—Biochemistry, medical sciences, and molecular, cellular and
developmental biology. It was not hard, in 1986, to predict this
reintegration, given my successful prediction on the progress of
genomics. I foresaw the ability to use computer modeling to design
and create new molecules with chosen functions. But I was a bit
optimistic in seeing the ability of genomics to tell us about the
biological locus of instinctive behavior.
—Geophysics, meteorology, oceanography and paleontology. The
first three have merged into planetary and earth sciences on the one
hand and climate sciences on the other. Indeed, the concern about
global climate change and sustainability has accelerated the study
of the interrelations of oceans and atmospheres, and paleontology
has proved a vital source of information (if ice cores are
considered paleontology). I could add geography to the list, given
the importance of studies of human habitation, energy use and
technology development in the issue of sustainability.
Third, specialization and reintegration still
compete. It was easy to foresee the dark side of
multidisciplinary studies—the claim scholars might make to
mastery of a broad interdisciplinary area without mastering any of
its constituent disciplines. This would make peer review and tenure
evaluations very difficult and controversial. One desperate hope was
of course doomed to failure—my dream that the National Science
Foundation and National Institutes of Health would stop trying to
predict the work that deserved funding, and instead reward those who
proved their work was worthwhile. For mature scientists that should
be easy. For young scientists I proposed that with every grant to a
mature scientist (based on her record) the university would receive
an additional 25 percent to be used for funding young scientists
within three years of a Ph.D. The universities would choose the
awardees. Back in the 1980s when I chaired the National Science
Board, I had already pushed for an additional mechanism: grants to
be made to young investigators by program officers, made without
peer review. The work of the awardees would be reviewed three years
later and the rating put in the program officer's performance file.
I proposed that an idea Herb Simon espoused back in 1985 would be
widely adopted in universities. The "Simon Standard" was
quite simple: "No one was allowed to publish pan-disciplinary
pronouncements until they had published at least one solid paper in
each of the disciplines drawn upon" (my words, not his). I used
Picasso as my model. When he was a teenager he showed he could paint
like Leonardo; he earned the right to represent a bill with five
lines on a piece of paper (which would sell later for millions).
Under this standard, the mono-disciplinary departments would survive
as keepers of the tools and standards in specialized areas of science.
Fourth, experiment and theory would be come increasingly
indistinguishable. On the data side it seems obvious
that when the quantities of computer-acquired data explode, one
builds algorithms into the computer analysis so the experimenter is
not seeing the output from individual sensors but rather a processed
flow of data which we should call metadata. The algorithms used to
process and simplify the data are themselves based on some theory in
which one has confidence. But the result is really not experiment,
independent of theory. For example, if you use symmetry properties
of viruses in the analysis of x-ray crystallographic images, the
result is clearly not independent of those theoretical assumptions.
The traditional separation of students into "theorists"
and "experimentalists" is no longer tolerable in 2003,
much less will it be tolerable in 2006. A similar line of argument
applies to the use of computation to exhaustion as a means for
"proving" mathematical theorems.
» Post Comment