Artificial intelligence system debates with humans

IBM has introduced the world’s first artificial intelligence system, which can debate with humans on various issues.

The IBM Debater, as it is called, can open a dialogue on about 100 different topics. After making a quadruple introduction to the subject under discussion, the “smart” machine listens to the arguments of his interlocutor, resolves them with a new four-minute placement, and in the end makes a “closing”, presenting for two minutes briefly his own position.

A month ago, Google introduced the Duplex system that can call a restaurant and make a reservation or book an appointment with a hairdresser. Systems such as IBM Debater and Google Duplex are important steps to make computers capable of naturally capturing people. A natural one that does not understand the interlocutor that it has to do with a machine rather than a man (this, moreover, is the basic criterion of success, which has been the British pioneer of the “smart” machines since the 1950s Alan Turing).

IBM has demonstrated the new system in San Francisco, with an Israeli student champion “dibete” rivaling, Nevada. The subject of man-machine dialogue was whether there should be government subsidies for exploring space (not just the easiest or most interesting topic for discussion, but that was the choice of IBM).

IBM Debater argued in favor of subsidies and Obama v. A quick “gallop” among those who watched the dibee showed that most people felt that the machine was better enriching their concerns over this issue than its counterpart. There followed a second machine dive with another Israeli, Dan Zaffrat, on whether or not the use of telemedicine should be increased (the computer was in favor while the man was against).

Based on the trials so far, as IBM researcher Noam Solon reported, the system – which has been developing for six years under the head of the company’s research laboratory in Haifa, Israel – has great room for improvement. Sometimes he makes mistakes while he manages to make meaningful dialogue on these 100 issues about 40% of the time.

In some ways, Debater is IBM’s new ambitious program of artificial intelligence after the Deep Blue computer that had won champion Gary Kasparov in 1997 and the Watson computer that defeated Jeopardy’s 2011 knowledge game champions.

In recent years, artificial intelligence systems have become increasingly capable of recognizing people, images and objects, or translating. But understanding the natural human language – something that is necessary in a dialogue – is a much more challenging challenge.