When computer scientist Feng-hsiung Hsu started his graduate work at Carnegie Mellon University in Pittsburgh in 1985, his goal was simple: to devise the best chess- playing computer ever.
His first attempt, a chess-playing processor called ChipTest, was controlled by a Sun- 3/160 workstation and could search around 50,000 moves per second. That might seem a lot of moves, but ChipTest skulked away from its first computerised chess contest with just two wins, one draw and two losses.
But by the time he was hired by IBM in 1989, Hsu’s work had become rather more ambitious. The Deep Thought project, as it became, sought to create a computer that could beat a grand master.
When Deep Blue – the successor to Deep Thought – beat Gary Kasparov in 1997 it was capable of exploring 100 million possible chess positions in a second.
Deep Blue was not just a vanity project for IBM. On the way to building the system, researchers made a number of breakthroughs in massively parallel processing – distributing workloads across many processors.
A similar man-vs-machine competition was a proving ground for Deep Blue’s progeny, natural-language processing system Watson. In early 2011, Watson took on champions at the US game show Jeopardy!– and trounced them.
Since winning at Jeopardy!, Watson has been putting its brain power to use at New York’s Memorial Sloan-Kettering Cancer Center. Here its ability to explore human language, combined with its number crunching prowess, has been used to analyse millions of pages of patient records, scouring the unstructured text to help doctors make diagnoses and answer treatment-related questions.
IBM believes that Watson has further uses in financial services, or even as a cloud-based service to general users. This is an impressive record for a technology that began life as a mediocre chess player.
See also: IBM Watson improves ability to understand the language of business
