Since IBM's victory over Garry Kasparov in chess with Deep Blue in 1997, the company was up to a new challenge. In 2004, Charles Lickel, IBM's Research Manager was dining with his co-workers and noticed that the restaurant they were in was too quiet.
He realized that the cause of the silence was Ken Jennings who was in the middle of his successful 74-game run on Jeopardy!. About everyone in the restaurant was watching the television between their meals. This sparked Lickel an idea. Lickel passed the idea to challenge IBM to the company's executive Paul Horn.
At the time, IBM's AI team was looking for a more complex challenges than chess.
Watson's predecessor was named Piquant. The AI was able to correctly answer 35 percent of clues, often requiring minutes to respond. But to compete in Jeopardy!, IBM needs only a few seconds.
The initial test for Watson was made in 2006 by David Ferrucci, the senior manager of IBM's Semantic Analysis and Integration department. Watson was given 500 clues from past Jeopardy! programs. While the best real-life competitors buzzed in half the time and responded correctly to as many as 95 percent of clues, Watson at first could only get 15 percent correct. In 2007, the team was given 3-5 years and a staff that consisted of 15 people to solve the problem. In 2011, Watson could finally defeat human constants.
The game ended with Ken Jennings with $4,800, Brad Rutter with $10,400, and Watson with $35,734.