IBM researcher Gerald Tesauro came to CSAIL Thursday evening to share IBM’s work teaching Watson superhuman strategies for its Jeopardy! challenge. Tesauro, who is widely known for his work developing a self-teaching neural network that learned to play backgammon, explained how IBM leveraged its Deep Blue question answering technology to make Watson a Jeopardy! champion
“This is the coolest project IBM has ever done,” said Tesauro, who delved into the various strategic aspects of Watson that helped maximize the system’s chances of winning Jeopardy!.