‘Computer science isn’t real science. You can tell that because they have to put the word science at the end of it, like they do with library science or social science – which are also not real sciences. Nobody talks about physics science, or chemistry science because those are real sciences.”
Those provocative words are not mine.
They were spoken by an Australian who works as a director of computer and communications infrastructure for the academic research community in his country.
I heard them this morning in a meeting room of the Banff Centre, at a group presentation called Cyberinfrastructure Uptake.
Cyberinfratstructure is a world made up expressly for the information technology summit I am currently attending.
Inelegant as the coinage is, it does serve the purpose of putting together two elements of information technology that are too often considered separately, both in the academic world and in the world of business: IT hardware (including computers, fiber-optic cables and routers) and IT software (just plain old computer programs).
The coinage, like the Australian’s words, may be more novel than informative, but both do serve to provoke some fresh thinking about computing and research and development – which is what this conference is all about, and why I am at it.
Our first morning speaker, if blessed with less Ozzy bluntness, had some interesting, big-picture things to say about the current accomplishments and lapses in information technology in the field of research.
According to his sources, the networked computer users in the world today currently generate enough data every 15 minutes to fill the entire US Library of Congress.
Furthermore, just considering science-based data, more such data is was created last year than has been created from the present back to when science began.
Herein lies all the hope, and all the trouble.
Though the value, utility and durability of all this data varies wildly, the resources needed to store, interpret and communicate it do not: bytes cost what they cost, whatever their inherent value as information.
On top of that, the huge informational infrastructure needed to generate, transmit and store all this stuff is, at present, uncoordinated, unstable and increasingly costly.
The cost of maintaining a big data warehouse – mostly in electricity to run and cool the computers, and in hardware to replace the parts that break down – is apparently increasing by a factor of eight with each passing year.
On the one hand, the existing and potential paybacks for all this investment and trouble are clear: Advances in physics, medical science and things like climate study have largely happened because of the shared information and expertise this high speed, high-capacity network makes possible.
On the other hand, the possible perils of all this activity, all loosely co-ordinated at best, is also great.
Though the whole point of creating this powerful infrastructure is to create co-operation and co-ordination between various researchers or research institutions, the truth is there is currently nothing in place to ensure that this kind of co-operation happens, or happens between the right people or happens in the right way.
Up to now, the emphasis has always been on increasing computer capacity, and pumping up the speed and reliability of the communication between them – pretty much a brute-force solution to all challenges.
Nothing like the same attention has been given to tailoring and developing the kinds of human partnerships using that network to maximize efficiencies and outcomes; and only slightly more attention has been paid to analyzing and revamping some of the geriatric software applications (many of them dating from the ‘70s and ‘80s) that run on all that high-cost, high-speed hardware.
In the old days, if a piece of computer code did not run very efficiently on existing computer hardware, nobody worried very much; it would probably run just fine when the new, faster processors came out.
As we reach the performance wall with single-process computing, however, that free lunch is starting to disappear.
Inefficient software is responsible for the vast majority of the vast amount of wasted electrical power generated annually by the academic research community.
As we enter the more “green” age of computing, the old, power-draining, brute force software solutions are going to become socially unacceptable because of their environmental cost.
Similarly, the benefits of all this high-speed, high cost computing has to become more evenly distributed around the research world.
The inconvenient truth under discussion at my second morning session (the one with the frank Ozzy) was that the huge investment made in cyberinfrastructure over the past several decades has actually benefited only a small majority of researchers, mostly in fields like physics, biology and chemistry.
The emphasis on big science and big-iron computing has led to the exclusion, or perhaps self-exclusion, of a whole range of research and researchers, who are either ignorant of the value all this infrastructure could have for them, or who ignore it because it has not been set up in a way or at a scale that meets their more modest needs.
Computing may or may not be a science in itself, but there is no question that its role in the new millennium is already huge, game-changing, and fated to be transformative.
It would just be nice if the computational and scientific worlds make sure they are on the same page about just what they are both going to transform into.
Rick Steele is a technology junkie
who lives in Whitehorse.