In the first ever public demonstration of an AI technology that IBM has been working on for the last 5 years, last week saw the tech giant’s “Project Debater” pit their artificial intelligence against two human debaters on the topics of government-subsidised space exploration, and telemedicine.
Although the computer had never studied the topics, it was able to present an opening argument by referencing a repository of newspaper articles and scientific journals which it had access to.
It then “listened” to the human counter-arguments on each topic, and presented a rebuttal.
You Win Some…
The outcome of the debates were judged based on before and after audience options on the topic, and while the argument for space exploration may have been a bit ambiguous in outcome, the IBM Project Debater was judged a clear winner on the topic of telemedicine, definitively converting 9 of the audience members to its point of view.
The project was started by an IBM research team based in Israel, not long after Watson, IBM’s leading AI experiment, beat two human quiz-masters at Jeopardy in 2011, and this demonstration follows Google’s Deep Mind / AlphaGo AI’s defeat of the Go world champion last year, a feat that most people believed we were at least a decade away from.
Not Always…Human
Although reporters said that the system showed its inhuman nature several times during the extempore speech which lasted for 20 minutes, IBM feels that it successfully demonstrated that AI could handle at least some of the complexities of human interaction, a much more complicated subject the rules of a game show or board game.
“Our life is not black or white. It’s ambiguous, it’s subjective,” said Ranit Aharonov, director of Project Debater. “AI will have to navigate in that territory.”
Clea Conner Chang, chief operating officer of Intelligence Squared Debates, added, “It’s amazing to see this technology pull from 300 million sources and distil it into what sounds like a conversational narrative in debate.”