.... bigger is not always better. When I did research at the Oak Ridge National Laboratory in the US a few years ago, managers used to say that supercomputers should be used in the upcoming field of biology-inspired research, as this would, in retrospect, justify the lab’s investment in a huge computing infrastructure. But they also worried that someone would get “lucky” and find a smaller, more tractable, model or a more efficient algorithm to solve the same problem on a laptop – making the lab’s investment into such projects superfluous.
Moving to Princeton University shortly afterwards, I realized that these worries were well founded.Top researchers, by asking the right questions and using clever models, could produce high-impact research results without ever needing supercomputers.Endres stresses (as I would) that supercomputers have a role in research but should be just viewed as a possible tool towards for scientific understanding, hopefully the weapon of last resort... They are just a means to an end. Furthermore, funding and politics must never determine scientific strategy.
The cartoon is courtesy of Bob Laughlin.
No comments:
Post a Comment