So How Does a Chess-Playing Computer Relate to "Real World" Applications?

IBM's Deep Blue:
Specialized Computing Potential for Technical, Commercial Applications

"The Chessboard is the World"   --; Huxley

For the first time in history, a computer has been designed that boasts enough pure processing power to actually pose a serious challenge to Garry Kasparov, the World Chess Champion. With the creation of the IBM Deep Blue* computer, IBM has designed a system that can search through a century of chess moves at speeds up to four hundred million positions per second. But why build a system that plays chess? Other than Kasparov, who would be interested in such a computer?

By learning from a "friendly" chess match -- an extremely complex and strategic game -- the computer playing against Kasparov may be programmed to solve complex but common problems which, historically, have been very costly in terms of both time and money. The technology developed by the Deep Blue experiment explores a new computing paradigm: combining both specialized software and hardware with general purpose machines to more effectively tackle problems. The power behind Deep Blue is an IBM RS/6000* SP* system, finely tuned with customized processor chips designed by IBM Research. This combination, in addition to expert knowledge, enables users to take on larger problems by analyzing a greater number of possible solutions. As a result, industries from express shipping and air transportation to health insurance, financial investment, cosmetics manufacturing and retail distribution could benefit from the Deep Blue system architecture.

Specifically, three commercial application areas have been chosen for study using IBM Deep Blue technology: Molecular Dynamics, Financial Risk Assessment and Decision Support.

Molecular Dynamics - Molecular dynamics is used within the pharmaceutical industry to discover and develop new drugs. Analyzing compound interactions on a molecular level requires massive amounts of processing and computational power. By creating a special purpose chip that can specifically address the complexities of molecular dynamics computing, a system can be designed which quickly and efficiently analyzes the interactions between atoms and molecules pertinent to the design of pharmaceutical compounds. The typical time necessary to bring a drug to market is 12 years and cost up to $12 million. A system based on Deep Blue technologies, specifically designed to target molecular dynamics, could drastically cut that development time to 6-8 years, bringing drugs to market faster at a lower cost, thereby potentially saving thousands of lives.

Financial Risk Assessment - Financial risk assessment is used in a wide range of applications to model trends in the equity marketplace. The applications consist of a wide variety of statistical methods that can be used to assess portfolio risk and predict future equity values. Typically, these modeling applications require a rapid assessment of value and risk associated with a large number of stocks and portfolios -- similar to the assessments which must be made in a chess game. Because the evaluations are largely independent of each other, parallel processing finds itself uniquely qualified for the job. As a result, a system such as the SP carries out these assessments with blazing speed, enabling users to visualize and capitalize on trends to offer a significant advantage in the marketplace.

Decision Support/Data Mining - Many business enterprises have accumulated massive databases, containing valuable information about products, sales, customers, and competitors. In addition to the information captured and framed by the database, there are also gems of hidden information which cannot be easily accessed via simple query facilities.

Data Mining is a process used to uncover hidden relationships and patterns in large databases. The process of discovery, however, requires a computer powerful enough to rapidly test thousand of hypotheses or data models. By processing data in parallel, in the blink of an eye the SP can run data mining software to verify hypotheses and highlight relationships in business data. Users may then leverage this insight to gain competitive advantage or to enter new markets, thereby providing valuable Decision Support.

Many types of businesses are using Decision Support today to solve difficult problems, including weather forecasting, marketing analysis and even in professional sports. The National Basketball Association uses Decision Support to search through millions of statistics and variables, helping a coach to put his most effective players in the game. By applying the same power which drives Deep Blue -- at one time only used in the scientific research community -- we can tackle an entirely new classes of business applications.

A few more examples of how companies have successfully applied Deep Blue technology to tackle their business challenges include:

• ShopKo Stores - This department store chain generates $2 billion in gross sales annually from 125 stores located in the Midwest and northwestern United States. Each store offers approximately 300,000 products for sale with a tracking code assigned to each. The result is a virtual warehouse of product data. ShopKo uses an RS/6000 SP to mine this data in an effort to better understand customer spending habits and patterns in inventory. The results enable store managers to more effectively merchandise store space, stock product and, what's more, plan and execute promotional and advertising campaigns. The analysis can be extremely specific: the weeks of the year when there's typically a run on pink Kleenex; how often shoppers buy greeting cards when they purchase red nail polish; or how many coffee makers should be in inventory during the second week of April.
• Argus Health Systems - A health maintenance organization, Argus is using data mining to improve the way it manages massive amounts of individual patient data. The company uses an SP to analyze huge amounts of data on patient drug treatments and their rate of efficacy. The goal is to assess drug performance and to better manage patient prescription benefits.