Boy Or Bot? Russian Robot Passes The Turing Test
In 1950, British mathematician, computer scientist, and philosopher Alan Turing wrote a paper entitled âComputing Machinery and Intelligenceâ in which he postulated whether or not computers would ever be able to think. However, since âthinkingâ is actually a surprisingly ambiguous term, his eventual line of inquiry led to a similar, yet much more specific question: Can a computer fool a human into think that it is also a human? While there are several variations on the test, the basic parameters remain the same: the computer must fool 30% of human interrogators after a series of five-minute chats. No computer has ever been able to pass the test — until now.
Meet Eugene Goostman, a 13-year-old boy for whom English is a second language. He is actually a computer, a Russian chatterbox program — but only 67% of the judges were able to determine this truth. With a final score of 33%, âGoostmanâ just barely managed to scrape past the threshold, generating an incredible amount of attention … and controversy.
There are those who believe that this particular incident shouldnât count, as the chatbox was specifically programmed to imitate a young boy struggling with language, baking a brilliantly devious explanation for awkward syntax and grammar goof-ups right into the core of the computer. One of the ways in which a judge might spot a potential chatbox is a contrast between whom the program claims to be, and how they speak. Many times, these programs attempt to mimic people of high intelligence, or at least those with a solid grasp on language. By purposefully muddying the linguistic waters, Goostman effectively veiled himself against an entire angle of detection, though one that trained judges would have likely seen through.
This brings up the second part of the controversy. Though the judges were all intelligent men and women, they were not specifically trained to recognize the telltale signs of a chatterbox program. While some think this detracts from the computerâs victory, others point out if a computer program were to be launched for cybercrime, telemarketing, or some other such purpose, those are exactly the kinds of people with whom it would interact — people who are largely intelligent, but not necessarily trained to identify or even expect this sort of digital trickery.
Despite initial jokes about Skynet and HAL 9000, the ramifications of this program are far less dire than some would have you believe. Goostmanâs programming, while tricky and deceitful, is still a far cry from an artificial intelligence. However, even though the win was through some tricky subterfuge and technicalities, it still merits some concern in regards to potential cyber criminals. If such programs can be written to fool the average Internet user, the potential for âautomated cybercrimeâ cannot be ignored. While Goostman itself isnât likely complex enough to pull off any sort of digital heist, the fact that this barrier was crossed at all signifies an interesting step forward in computation. The threshold has been passed. Whether you see it as blatant cheating or clever thinking, that much is certain.
Image Credit: Thinkstock