NetNews Usenet Archive 1992 #20/NN_1992_20.iso/spool/comp/os/research/925

Internet Message Format | 1992-09-14 | 6.9 KB

Path: sparky!uunet!spool.mu.edu!olivea!hal.com!darkstar.UCSC.EDU!osr
From: [e-mail] (Piercarlo Grandi)
Newsgroups: comp.os.research
Subject: Re: Future of CS & CE research (Petition)
Message-ID: <[e-mail]>
Date: 14 Sep 92 18:51:30 GMT
References: <[e-mail]> <[e-mail]>
Organization: Prifysgol Cymru, Aberystwyth
Lines: 122
Approved: [e-mail]
NNTP-Posting-Host: ftp.cse.ucsc.edu
Originator: osr@ftp

On 13 Sep 92 23:33:32 GMT, [e-mail] (Ramesh Bharadwaj) said:

rxb> The notion that research funding should be based solely on the
rxb> criterion of "concrete demonstration of benefit to the nation",
rxb> although regrettable because it may slow down basic research in the
rxb> short term, is a reform that will have a positive effect on the
rxb> whole -- a few good researchers may be cut off, but it is
rxb> undeniable that more good research than bad is going to get funded.

Maybe you are right, but the main focus, as I read it, of the complaint
was the question:

  Why should we be treated differently from any other subject?

rxb> I see it as a knee-jerk reaction to years of systematic
rxb> misrepresentation and irresponsibility on the part of many
rxb> researchers in "computer science", particularly the ones at
rxb> "prestigious" institutions.

But what in this obervation of yours what is specific to computer
science?  I think the same could be said of virtually all other
subjects, and in much the same degree, whether in the sciences,
engineering or the arts.

It can be reasonably argued that this is not a bug -- it is a feature.
Most basic research that gets funded nowadays is probably a waste of
money.  But basic research is not like a tidy garden, it's a compost
heap. Or maybe it is like dry wells in oil prospection; you have to
drill a dozen for every productive one you expect to find. The same goes
for drug discovery, where you test thousands of compounds each year
hoping to find one that works, or for marketing consumer products, where
only one in two dozen succeeds.

It gets quite easy then to say this is a waste and that one could save a
lot of money (several dozen million dollars for each well) and by
drilling only the wells that will not be dry, but not as easy to do it
in practice (if you know of any method to do it, and can give me a
concrete demonstration of its benefit, I will invest all my savings in
your company).

rxb> I think the corporate world, the general public and funding
rxb> agencies are sick and tired of tall claims made by researchers in
rxb> "A.I." or, for that matter, by "computing theorists" (whose
rxb> "results" often bear an uncanny resemblance to earlier work in
rxb> mathematics or logic -- they are often a rehash of previous work
rxb> couched in different terminology, published in a different journal,
rxb> and read by a different audience).  It should therefore not come as
rxb> a surprise that the funding agencies are beginning to ask where
rxb> their money is going.

There is an age old proverb about Ph.D.  dissertation, for example,
which characterizes them as "the shifting of old bones from one grave to
another". This adage predates computer science, and indeed computer
science... Such, more or less unfortunately, are the rules of the game.

rxb> Merging Computer Science and Computer Engineering is not such a bad
rxb> thing -- it will weed out the chaff and avoid unnecessary
rxb> duplication of effort.  To start with, the division is artificial
rxb> and has done more harm than good.  However, unsuccessful
rxb> Philosophers, Linguists, Physicists, Cognitive Psychologists,
rxb> Mathematicians and Logicians who have found a safe haven by calling
rxb> themselves computer scientists, may not find the move so welcome --
rxb> having to rub shoulders with real engineers may uncover
rxb> embarrassing gaps in their knowledge of computing.

Well, it might be argued that, broadly, Engineering is about the
application of known results to practical tasks, science is more about
the discovery or invention of new, and usually unexpected, results. It
seems rather hard to apply the same criteria for funding research in
such different areas.

Indeed while it may seem odd to an engineer to see Philosohers,
Linguists, Physicists, Cognitive Psychologists, Mathematicians and
Logicians dabbling in computer science, maybe one of them holds the key
to those more or less revolutionary novelties that research is all
about.

It is fairly clear that whenever there is "free" money slushing around,
a lot of it is going to be used in silly, pointless, or even dishonest
ways; but the tradition for all disciplines, and admittedly it may be
well self serving, is that it is better to fund a lot of silly research
for the sake of the few gems that more or less by luck spring up, than
to try to pick winners.

And the attempt to introduce pre-selection of winners by committee (or
worse) as required by "concrete demonstration of benefit to the nation",
is hardly going to work; as somebody is fond of remarking, even the
present system of peer review, whose aim is not to select the winners,
but merely to weed out the obvious losers, often leads to the situation
where you get funding only if the results of your research are reported
in detail when you apply for funds to initiate it.

Very serious people have proposed abandoning peer review altogether, and
any attempt at picking winners, and to use instead a lottery system!

The "concrete demonstration of benefit to the nation" gauge, which
sounds so appealingly simple, is terribly difficult to implement, is
arguably best left to the invisible and rather stochastic operation of
the marketplace, is bound to stifle basic research, and implies so many
moral hazards (already fairly evident in the peer review system); it is
little more than a catchy phrase.

All this is extremely apparent in even a narrow field as OS research;
clearly the easiest "concrete demonstration of benefit to the nation"
could be made about research on MSDOS memory management utilities, but
maybe that's a field best left to Qualitas and Quarterdeck; so we
continue to do demonstratably pointless research on capability and
microkernel based OSes that nobody will every give a damn about -- but
then who can be sure, given that even in the system software marketplace
the lead time from invention to widespread adoption is twenty years?


In the end the issue of being resigned to wasting most of the funds
invested in basic research, or else trying to pick winners in advance,
is going to be settled in the same way as for oil prospection or drug
discovery or consumer product marketing: those who can afford to pay the
high price of wasting most of their investments in failures waiting for
a huge winner do, and the others license the results and whine.
--
Piercarlo Grandi                   | JNET: [e-mail]
Dept of CS, University of Wales    | UUCP: ...!aber-cs!pcg
Penglais, Aberystwyth SY23 3BZ, UK | INET: [e-mail]