Economic Supercalculation

A fellow Numeriker once asked me this: given that an improved algorithm can cut back on your computation time considerably, would it not be wiser to spend research money on developing algorithms, instead of the current large investments in sports-hall sized, power-guzzling, maintainance requiring supercomputers.

I thought I'd revisit my answer in light of my reading of Human Action, by Ludwig von Mises. In my original answer I said that although the equivalence of algorithms and computation power is correct, we need to consider not only the means, but the ends: if the application research field in question (e.g. medicine, biology, cryptography) is considered important, then it shouldn't be forced to rely on the idea that progress in algorithms will come. It is a fundamental feature of science, including mathematics, that progress is not only a function of money, so if you want results now, it is more straightforward to go for the known technology of supercomputing – to just give the researchers processor time.

להמשיך לקרוא

מודעות פרסומת

The next step in scientific collaboration

Today I'm writing of something that may be of a more general interest, so I'm doing it in English, with the hope that It would save me work later in explaining this to others.

Science can confuse superficial thinkers: while Karl Marx, for example, claimed in The Capital that nothing can be both an individual activity and at the same time a social one, the natural science is exactly that. Research is most often done individually, with collaboration taking the form of division of labour among groups; but scientists crucially depend on the social environment. Without the dissemination of new knowledge, the learning of new methods, and the attention to the work done by others, a scientist would be left to wallow in irrelevance. Worse, he might waste his time repeating work that has already been done.

The importance of the social side of science has driven development of social tools for science, alongside with the development of technical aids to research. The first such tool was the post, his majesty's royal mail carriage, through which individual scientists shared their recent findings with like-minded naturalists. With the growth of scientific traffic in the mail, groups such as the Royal Society and others began collecting articles of interest and publishing them, using the already-old technique of printing, to a wider audience. Thus was born the scientific journal.

In the passing decades the journal has been improved and perfected, and has even made the move to the series of tubes passably well (although much room for improvement exists). Then came other tools of collaboration: the train, the automobile, the airplane, the phone, the email – all of which help scientists keep in touch, be aware of the focal points, trends and open questions in their field, and learn of new technologies or knowledge that comes about.

With all this panoply of collaboration tools, one would be excused for thinking that duplication of work and ignorance of tools of the trade are a thing of the past; yet it is not. Science is undergoing a transformation in the computer age, which has left many scientists out of touch and badly equipped. להמשיך לקרוא