03/06/2012

Why You Should Care About Supercomputers

Later this year at Oak Ridge National Laboratory the world’s fastest supercomputer will be turned on. Even for those of us that get excited about multi-core processors, gigabytes of RAM, and superfast video graphics, thinking about supercomputers is just sort of “out-there”. After all, most of us won’t ever have access to one. Most of us don’t even know what they actually do on those computers. Who really needs all that computing power anyway?

But supercomputers have, and will impact our lives.

Several decades ago scientists embarked on a mission to map the entire human genome – that is to identify all the genes in human DNA. Originally planned to take 15 years, the project came to completion in just 13 years due to increases in technology and supercomputing. The human genome project (HGP) completed in 2003 will ultimately help diagnose, treat, and prevent disorders. Already parents are offered the option for certain genetic tests while their baby is still in the womb. Other tests are available for detecting diseases such as Huntington's or Alzheimer's.

Depending on what you read, the 2008 housing and stock market crash can be blamed on supercomputers – at least in part. You may not realize it, but if you invest in a fund on the New York Stock Exchange (NYSE) you may be placing your financial future in the hands of a computer. Experts say 73% of trades on the NYSE are placed by computers. Now this doesn’t mean a broker or trader puts in an order that the computer processes it. No. It means that the computer analyzed the data and decided whether to buy or sell without any intervention from a human. This happens today, thousands of times every day.

Starting to get the feeling these supercomputer thingies might just be important after all? It gets better.

You probably know that last year IBM’s supercomputer, Watson, beat two Jeopardy champions in a tournament over three days. Total winnings for Watson topped $77,000, while the closest human competitor only reached $24,000. Watson is a new type of supercomputer, designed to be super-fast at processing data, finding patterns, and drawing conclusions based on probabilities. With the equivalent processing power of 1 million books per second, the humans really didn’t have much of a chance. But sheer processing power alone isn’t what made Watson so successful. Its rudimentary cognitive abilities (recognizing patterns) allowed it to do more than just search the database for an answer. It actually tried to figure it out. Not exactly like human thinking, but it is a huge step closer.

So close in fact that IBM is working with several organizations that want to bring Watson into the examination room – at your doctor’s. The idea is that Watson can consider all the case histories and known medical information to make a better, more informed diagnosis than your doctor can with limited knowledge (compared to Watson). The idea seems to be catching on, but where it will go no one knows yet.

Now before you get all worked up about Skynet becoming a reality, just keep this in mind: For the moment these computers are still programmed by humans. They just do what they are told – even though that can be very impressive.

The next step of course is to create something even more human-like in its ability to process and learn. IBM is working on “cognitive computing chips” that are totally different than today’s processors. They aren’t designed to be programed. They will learn, remember and be able to correlate information. Where will that take us? Science fiction seems to be quickly becoming a reality, but which fiction, or future it will be we’ll just have to wait and see. Personally I think the science fiction versions of the future of supercomputers are much more about us than them. I’m hopeful that if we ever do create a self-aware computer it will not mimic our hostility or self-centeredness. Perhaps they will set the example for us.

What is your favorite supercomputer villain? Can you name one well-known movie or book where the supercomputer didn’t want to kill or control all humans?