In autumn 2004, the company was looking for a new challenge to follow 1997’s defeat of chess champion Garry Kasparov by Deep Blue. As one story goes, IBM Research manager Charles Lickel was at a steakhouse near IBM headquarters when the dining room emptied so customers could crowd around a television in the bar and see whether Ken Jennings’ streak of Jeopardy! wins would continue. Soon, Lickel began pushing the idea to build a computer that could beat Jennings.
It would be much harder for a machine to win at Jeopardy! than at chess, and many in the artificial intelligence community–inside and outside IBM–doubted it was possible. They believed that only the human brain could understand the concepts, ideas and linguistic nuance wrapped up in the game’s clues. But the ability of new algorithms to decipher human language and analyze vast stores of unstructured data in a few seconds gave Watson a chance, and his creators made the most of it.
Jennings and Watson’s other human opponent, Brad Rutter, were unable to match the machine’s question-answering horsepower, but Jennings insisted in 2013 that there is still a lot to be said for “good old-fashioned human knowledge.”