IBM began developing artificial intelligence for the debates in 2014, as a separate branch of the development of “computer thinking”, as opposed to systems with accurate calculations. Computer Deep Blue, who in 1997 defeated Garry Kasparov in a game of chess, acted in a strict field of rules. AlphaGo won the game in go where the winning strategy is blurred in infinity of options. And on June 17, 2018, the Project Debater system defeated a man in a verbal dispute!
Artificial Intelligence Project Debater is designed to implement a historically complex goal – filtering sources of information according to their value, based on the probability of disinformation among them, deliberate or accidental. Simply put, he knows how to reason, soberly assess the arguments and pick up the counterargument, in order to reveal the information necessary to achieve the truth. Only at the same time it is deprived of the influence of emotions, bias, ambiguity and generally does not have its own opinion as such.
Opponent Project Debater in the indicative test-debate was made by Noah Ovadia, champion of the national debate in Israel. AI did not have access to the Internet, but there was a knowledge base of hundreds of millions of scientific publications. The topic “Public funding for space exploration” was chosen at random from 100 options, AI gave the command to fight for, Ovadia opposed. And I lost, according to the cumulative assessment of the audience.
AI acted prudently and meanly – he agreed with most of the opponent’s arguments, but immediately argued that their significance was not so great as to close the question. As a result, Ovadia just ran out of arguments and time, and the AI showed that it can draw enemy weapons against it.