UConn students had the opportunity to watch history in the making during a streamed broadcast of “Jeopardy!” Wednesday night. The episode featured one peculiarity: human contestants battled it out with a computer.
IBM created an artificial intelligence program named Watson and designed it to answer questions posed in common speech. In 2006, Watson only answered about 15 percent of questions correctly during initial tests of “Jeopardy!” but by 2008, Watson could compete with the hard-hitters of trivia. In 2010, Watson regularly bested human contestants.
Monday night premiered the faceoff between Watson and Ken Jennings – winner of 74 “Jeopardy” matches in 2004 – along with contestant Brad Ritter. The night ended with Watson tied with Rutter at $5,000 each. But, by Tuesday night’s end Watson had a lead on Jennings of more than $30,000. Not bad, considering Watson had a modest 15 to 20 percent confidence level on most questions.
But Watson’s performance last night astounded audience members: the room-size computer triumphed over the humans with $77,147 to Jennings’ $24,000 and Rutter’s $21,600. Along with first place came a $1 million prize, all of which will be donated by IBM to World Vision, an evangelical relief and development organization.
Watson, in a non-televised match of “Jeopardy!” defeated reputable science correspondent Miles O’Brien, who stated that “it wasn’t elementary, it was really ugly,” when reflecting on his defeat.
In attendance last night was Sharon Nunes, vice president of green innovations for IBM.
“We hope you’ll develop a better appreciation for advancements going on in computer science,” Nunes said. “Just think, we could have this information in our cell phone in maybe ten years.”
“It’s inspiring – it makes me look forward to doing something with this type of technology in a career,” said Dave Powell, a 4th-semester cognitive science major. “It’s very interesting and I want to research this more.”
Using 15 terabytes of RAM and a complex web consisting of thousands of algorithms simultaneously, Watson analyzed the question asked and sifts through references like bibles, novels, plays and encyclopedias to arrive at the answer. The more algorithms that affirm the same answer, the higher the likelihood it is correct.
But in a game full of wordplay and puns like “Jeopardy!,” an automatic search engine alone will not suffice. The challenge is Watson’s understanding of culturally-nuanced questions inherent in the game. This language barrier exemplified the challenge faced by all of Watson’s programmers.
“I’m really impressed by the breadth of information that the computer can access and the speed. It’s fantastic, seeing so many real life applications,” said Ashley Calder, a 4th-semester computer science major.
The event was sponsored by CLAS, the Honors Program, the Neag School of Education, the School of Business, School of Engineering, School of Nursing, School of Fine Arts, the School of Pharmacy and UITS.