YORKTOWN HEIGHTS, N.Y.-"Watson" isn't brushing up on any trivia today.
Watson, now maybe the world's most well-known supercomputer (with its ownNew York Times Magazine profile to prove it), is half hidden behind a black curtain and stationed in the mock "Jeopardy" studio that IBM has built inside its massive T.J. Watson Research Facility here.
During a recent visit there, IBM engineers could be seen tweaking Watson, which, though based on some of the same technology as the company's other Blue Gene supercomputers, utilizes natural-language capabilities. Although Big Blue is still testing Watson against "live" Jeopardy contestants, the supercomputer is expected to appear on the real game show sometime later this year.
While Watson, much like its chess-playing predecessor, Deep Blue, is an impressive way for IBM to capture the public's imagination, the supercomputer is only a small part of what goes on at the Watson facilities, both here and in neighboring Hawthorne, N.Y.
What is really driving a small but growing part of IBM's research these days is analytics, which is essentially a discipline that uses mathematics and data to arrive at the most optimal decision. For years, IBM research and math departments have played a key role in the company's research; now the company tapped these scientists and researchers to be part of its massive Global Services Division.
One day, in fact, Watson could find itself, rather than on TV, part of this very same analytics push--specifically focusing on health care, where its language software and massive computational ability could assist in the areas such as molecular biology and genomic sequencing. As part of this direction, IBM is looking to increase the number of people in its math department from about 110 to 200 or so, especially concentrating on those with expertise in very specific fields of study, including operations research, machine learning and statistics.
"We are looking for special people. We don't want people in the ivory tower," said Baruch Schieber, manager of Business Analytics and Optimization at IBM Research. In other words, he is looking for mathematicians interested in real-world work. "We want people to get in here and understand how people manufacture and how companies distribute their goods from the warehouse to the supermarket."
Large IT companies, from IBM and Oracle to Hewlett-Packard and Dell, all have services divisions, which are becoming increasingly competitive and lucrative. In a recent report, IDC found that global IT-services spending is slated to grow from $573.4 billion this year to $685 billion by 2014. When IBM announced its financial 2010 second-quarter results, the services division reported a 2 percent increase in revenue, to $13.7 billion.
IBM's approach to IT services is somewhat different from that of other companies. It is able to tap into its own massive research division and incorporate in-house scientific research, such as analytics, into a business division such as Global Services. For the past year, IBM has highlighted this approach in its Smarter Planet initiative, but the concept actually goes back to 2002, when IBM bought PricewaterhouseCoopers. This began a nearly 10-year effort to bring more research and analytics into the services division.
Schieber explained that, though IBM has been talking up its particular approach to IT services for the past few years, applying analytics to business problems actually goes back about 20 years, to when IBM worked with American Airlines on a customized fleet-and-crew scheduling project.
IBM researchers work within two different disciplines of analytics, which then come together. The first is business analytics, which uses data to make predictions and forecasts; the second is optimization, which combines predictions and business constraints to produce an outcome and plan to meet differing objectives.
A recent example of this cooperative process is a project with the New York State Department of Taxation to collect tax revenue owed to the state.
First IBM research looked at the situation, researched the historical data and applied both types of analytical methods to the problem. The company then came up with a combination of software-including data warehouse, database management, and business intelligence-to determine which people would respond to a simple letter from the tax division and which ones would require a full-scale audit.