THE FASTEST COMPUTER
MANILA, November 18, 2004 (STAR) DE RERUM NATURA By Maria Isabel Garcia - The fastest known physical action of a living thing is that of a small insect called a midge. It beats its wings at a rate of 57,000 times per minute or 950 per second. But when it gets nervous, this number can rise to over 2,000 beats per second. In Fred Flintstone’s Bedrock, the midge’s nervous wing-flapping may be its best bet at automating its world to the max. In our current world of nanotechnology and supercomputers, we have reached the tera-speed, i.e., in the trillions of calculations per second (teraflops), from the first supercomputer in 1976 called Cray 1, which was capable of only 80 million calculations per second (megaflops). IBM has just announced that it has achieved for its Blue Gene/L computer the speed of 70.76 teraflops. Blue Gene/L is said to be only a hundredth of the size of and requiring only 28th of the power per computation of Japan’s Earth Simulator, the former record-holder. IBM intends for Blue Gene/L to even top its own current prototype speed by next year.
Supercomputers are designed to handle massive amounts of information for specific purposes. The other supercomputers that ranked top in terms of speed are Columbia, USA, 51.87 Tflops; Earth Simulator, Japan, 35.86 TFlops; MareNostrum, Spain, 20.53 Tflops; and Thunder, USA, 19.94 Tflops. Speed technology advances at such a rate that the supercomputers are ranked every six months. This Blue Gene/L is envisioned by the US Department of Energy to handle information on nuclear stockpile. Other press releases say that it would also have wider implications not just in the energy sector but in biotechnology as well. Other supercomputers mentioned handle information involving climate change.
I want to know what "speed" means in a larger context. Impressive as this 70+ teraflop is that the Blue Gene/L is capable of, I want to be able to put it in perspective so that it will be meaningful to someone like me who will never work with a supercomputer but whose life, together with the rest in this planet, will probably depend on one, considering that it will control and test nuclear stockpile as well as make sense of the human genome and the gene computations necessary for medical and other scientific purposes. Hans Moravec, a research scientist at the Robotics Institute of Carnegie Mellon University, devotes a lot of his time computing the "flops" that living things are capable of processing that helps him develop computers/robots that could aid humans in specific tasks. He writes in the book The New Humanists (John Brockman, ed. Barnes and Noble Books: NY, 2003) that current PCs today still only approximate the computing power of the milligram-heavy nervous systems of insects or small vertebrates like that of a dwarf goby-fish. Moravec also computed the retinal to brain calculations that are involved when humans see images and it is in the order of around 100 teraflops, compared to the Blue Gene/L’s current 70+ teraflops. That is just in terms of visual imaging. What about the computations involving the other senses? And what about their convergence in "synesthesia," when all the senses produce a sensation that is more than just the sum of the different inputs to the various senses? Moravec said sensory imaging is one of the deepest levels that still elude the simulation of the best computers we have today. Speed technology can help us diagnose the condition of such complicated matters like nuclear stockpile but we still have not come up with an artificial intelligence that can recognize the many expressions of the human face or even recharge itself by finding a plug in the hallway of the robotic laboratories. But perhaps it is because it is not speed that is required to achieve these things. Somehow, I am apprehensive that the speed at which the Blue Gene/L or any other supercomputer, wherever it is, tasked to do diagnostics on nuclear stockpile or any other action on it, may miss out on some things important to ask which the scientists did not find relevant to include in the list of diagnostic computations/questions. Questions along the lines of: "Are you sure you want to eliminate this portion of the planet, together with all the other living things in it, now?" And then maybe if the answer is "yes," maybe the computer could still ask: "Do you already know how many people there are in this portion of the planet you want to erase and do you also already know how many flora and fauna find home in it as well?" And then if the answer were still "yes," I would prefer that the computer then ask random questions about each and every living creature there is in that place in question. Because I doubt if the scientists could answer all since we really do not know everything there is to discover about anything in the world. Why do you think we are still here doing science? I can imagine IBM scientists seriously frowning at me (and my sanity) and showing me out of their laboratories if I ever suggest those questions.
I wonder if they showed Dr. Lewis Thomas the door since he suggested something similar in his time, even calling it an "earnest proposal." Dr. Lewis Thomas is one of my favorite science writers who wrote not just in medical journals but also in The New Yorker, the Scientific American and other good science publications for non-scientists. He was a prominent Harvard American medical doctor who, among other positions, headed several research departments of universities and served as president of the memorial Sloan-Kettering Cancer Center in New York before he died in 1993. He wrote books like The Medusa and the Snail and The Lives of a Cell. He was a medical doctor but wrote beautifully. "But" because I have the utmost respect for medical doctors yet find it so rare among their kind that they be able to write unless it is in passive sentences and with torturous jargon. But they, of course, have more important things to worry about than clear, elegant sentence construction so let me just focus on the rarity who was Dr. Thomas and what he earnestly proposed. He wrote in The Lives of a Cell (Penguin Books, NY, 1978) that since nuclear scientists would not really want to overlook anything while calculating "acceptable levels of megadeath," they may just want "to defer action until maybe we already have one set of complete information about one single living thing." For this, he even nominated a protozoan, Myxotricha paradoxa, that lives in the digestive tract of Australian termites. I personally would have started with the sloth since I want to be able to differentiate it with finality from some humans I know. Dr. Thomas predicted that the supercomputer might end up saying, "Request more data." And until then, "do not fire." I predict the same. I hope that tera-speed computers not just hasten the process to realize our option to act but also offer us our option to restrain.
* * *
For comments, e-mail firstname.lastname@example.org
Reported by: Sol Jose Vanzi
© Copyright, 2004
by PHILIPPINE HEADLINE NEWS ONLINE
All rights reserved
PHILIPPINE HEADLINE NEWS ONLINE [PHNO] WEBSITE