Russian scientists one step from creating AI?
All the contesting programs had to undergo the so-called Turing test. In 1950, UK mathematician Alan Turing, one of the first developers of computer technologies, suggested a test (although, at that time, it had little scientific significance and was rather a game). The rules are simple – the examiner communicates with an anonymous partner who can be either a human or a computer program. The examiner neither sees nor hears his partner – he only receives printed answers to his questions from his hypothetical interlocutor. The examiner has to guess whether he is communicating with a real human or with a computer program. As a rule, after some time, the examiner guesses it right, because no computer program that can fully imitate human thinking has been created yet – if it may ever be created at all.
In an interview with the Voice of Russia, Mikhail Gorbunov-Posadov from the Institute of Applied Mathematics said:
“Initially, the programs that were tested by the Turing test “fooled” the examiner by a number of tricks. Pretending to be a human, the computer “partner” asked to repeat the question as if it didn’t hear it properly, used grammatically incorrect phrases and the like. However, such a program may “fool” an examiner only for some time. Rather soon, the examiner did guess that his partner was not a real human.”
“In fact, initially, programs that imitated human intellect had little practical use. It won’t be an exaggeration to say that they still have little practical use right now. Still, this doesn’t at all mean that attempts to create artificial intelligence are nothing but a fruitless game.”
“To make a computer device imitate human speech as close as possible, creators of computer programs have to hold very thorough linguistic studies of real humans. I believe, one should never call any scientific experiments useless,” Mr. Gorbunov-Posadov says. “Sometimes, scientific investigations may bring results unexpected by the scientists themselves.”
It is believed that a real artificial brain (which, as said above, has not been created yet) would be able to “fool” the examiner in answering not less than 30% of the questions in the Turing test. The Russian program called “Evgeny”, which took part in the contest in London, passed the test, that consisted of 150 questions, with the result of 29.2% - very close to the “desired” 30%.
Does this mean that Russian scientists are now one step from creating a real artificial intelligence? Well, at least, they are convinced that very soon, they will be able to create computer programs which will look for information on the Web in a more “intelligent” way than the already existing search programs.
Mikhail Gorbunov-Posadov says:
“The current search programs can only look for a word or a group of words on the Web without understanding the meaning of these words. The new programs, of course, won’t understand the meaning of the words like people do, but they will also take certain aspects of their meaning into account, which will make the search more fruitful.”
“It is probably too early to label the Russian creation artificial intelligence,” Mr. Gorbunov-Posadov concludes. “But what the Russian scientists did is undoubtedly a serious breakthrough. For all the imperfections that their creation still holds, it can already cope with many problems that used to be unsolvable for computer programs of the earlier generations.”