This weekend, at the headquarters of the Royal Society of London, has been a significant achievement that has made great headlines in the technology and mainstream press. Finally, after several times when it was about to do, a machine has been able to overcome the celebrated Turing test. A test on the original idea of the mathematician Alan Turing, would point out that a computer is “intelligent”. Is it a landmark? Yes, especially the historical relevance of the proposed Turing. Does this mean that today’s computers are smarter than a week ago? Not at all.
Who has managed to overcome the challenge is a conversational robot named Eugene Goostman, who pretends to be a child of Ukrainian origin for 13 years, and that has been developed by three Russian computers. It is not the most ready supercomputer in the world, or a computer of last generation able to eat him all the pawns to Magnus Carlsen. It is a simple program, a simulation created to make it appear that there is something intelligent on the other side: he is not intelligent, but that its developers have been able to achieve to imitate a person.
And imitate a person who is not particularly intelligent, and here’s the trick: is a child whose native language is not English. Therefore, cannot be expected to know much, nor that to speak particularly well. “We spend much time developing a character with a credible personality. Our main idea was that he can ensure that he know something, but because of his age, it is also perfectly reasonable that he will know everything”, said Vladimir Veselov, as recorded by a note from the University of Reading, organizer of the event.
Eugene passed the Turing test after 33% confuse a jury of experts in artificial intelligence and people without specific training in the art, chat conversations for five minutes. The minimum required to pass the test is 30% confuse jurors and to this end, developed this and other programs that had been about to do before. The Eugene himself stood at 29% a couple of years ago and in 2010 Suzette had managed to deceive, for the first time, one of the judges, earning the Loebner Prize, a competition organized Turing test annually.
“This year we have improved the controller box, which makes the conversation much more similar to that of humans than programs that merely serve to answer the questions”, explains Veselov, one of the creators of the program and specialist Amazon in semantic encoding. This type of chatterbots, as it is called in English, is the type of software used on commercial websites to help customers with their purchases.
Manuel de Leon, the director of the Institute of Mathematics (ICMAT-CSIC) Science, explains that the real significance of the achievement is scarce. As explained Matter, believed that if a Turing machine behaves in all respects as smart, then it must be smart. “However, the development of artificial intelligence does not move that way”, he explains. When the British mathematician wrote his text Computing Machinery and Intelligence in 1950 was like putting the asphalt road to the first major artificial intelligence. 64 years later, we have built highways and freeways much more sophisticated, and the Turing test is a dead end to development of these technologies.
“This victory does not mean no progress, is not going to change things. It is more the iconic, especially now that we have celebrated the year of Turing, the social impact and media”, highlights De León, to conclude by recalling that in your environment it did not cause the slightest stir: “I am a member of the Executive Committee of the international mathematical Union, and whenever something interesting happens it took nothing to cross us comments. This happened on Saturday and no one has said anything”, he continued.
The importance is not, therefore, has been made a breakthrough in the field of artificial intelligence, but in the symbolic: the boundary between human and robot is becoming increasingly difficult to draw in a world that already prevails Siri, the wizard of Apple devices, and Watson, a supercomputer capable of defeating people in game shows using irony. It had also passed the Turing test in other areas, as in video games, getting bots seem human players. For greater symbolism, Eugene achieved his feat on June 7, just six decades after the death of Turing.
The professor of cybernetics at Reading University and head of the historic event on Saturday, Kevin Warnick, said “of course, the test has implications for society today. Owning a computer can fool someone into thinking it is a trusted person is a real wake-up call on cybercrime”.
“The Turing test is a vital tool to combat this threat”, he argues. But, as says the expert Ramón López de Mántaras, “artificial intelligence is already everywhere” and try to imitate the human “is absurd”.