The New IBM Supercomputer Beat the Humans in the Debates

The challenges between human beings and computers are an old and quite disturbing obsession with IBM, at least since the time (it was 1996) in which the chess champion Garry Kasparov was defeated by the IBM Deep Blue computer. But if the idea that a computer is stronger than a human being in a game based essentially on computational skills like chess, far more surprising,  or if you prefer disturbing  is to think that a machine can compete in a human activity in the broader sense of the term as the debate.

Yet more or less this happened on Monday, June 18 in San Francisco. An IBM system called Project Debater participated in a debate with two human participants, and it did not do badly. Indeed, in the public opinion he won one of the two debates. The rules were similar to the traditional ones of the competitive debate: each participant had 4 minutes to present his arguments, then 4 others to answer, and 2 for the final arguments.

Project Debater gave his best when he showed he knew how to predict the possible arguments of the opponents and challenged them in advance, and when he even showed a certain sense of humor saying “I would boil the blood in my veins, if I had “. In other circumstances, he has shown that he still needs some fine-tuning, as when he cited data (a ruling by the Supreme Court of Iowa) that had little or nothing to do with the subject discussed .

“We are not obsessed with winning and losing,” said Dario Gil, vice president of Artificial Intelligence at IBM Research. “We are more interested in seeing Artificial Intelligence struggling with human ambiguity, reality, and context.” We want to see if it is able to write well, be persuasive, and build good arguments. ” In short, to behave like a human being.