Alan Turing: Are machines thinking yet?
Google is honoring war hero and artificial intelligence pioneer Alan Turing on his 100th birthday Saturday with a Turing Machine-themed doodle. How close are we to Turing's predictions?
In 2006, Dr. Robert Epstein went online in search of love. The Harvard-educated psychologist joined an online dating service, and soon after he began exchanging emails with someone he described in Scientific American: Mind as "a slim, attractive brunette."
"Ivana" said she was from Russia, which would help explain her idiosyncratic English. But there was something else that was odd about her prose. Her emails, while warm and affectionate, came across as "a bit redundant and, let's say, narrow in scope," wrote Epstein.
After four months of online correspondence, Epstein, a former editor-in-chief of Psychology Today who has written extensively on love and relationships, had become suspicious. Ivana, while responsive, tended to steer away from specifics, he noticed. She would also often ignore direct queries.
He eventually tested Ivana by sending her a couple lines of gibberish. She responded, apparently unfazed, with a story about her mother.
Epstein then realized that, after all these months, he had been romancing a piece of software.
None of this would have surprised Alan Turing. The British mathematician and computer scientist, whose 100th birthday Saturday is commemorated on Google's home page, predicted in 1950 that, about five decades hence, computers will have advanced so far that, 70 percent of the time, an average person would be unable to distinguish a computer from a human after five minutes of continually exchanging messages.
He was not far off. In September last year, a web application called Cleverbot took the Turing test alongside humans. Of the 1,334 people who conversed with Cleverbot for four minutes, some 59 percent judged it to be human. (Strangely enough, the actual human participants were thought to be human only 63 percent of the time.)
Cleverbot's creator, British programmer Rollo Carpenter, is also the two-time winner of the Loebner Prize, an annual contest begun in 1990 that awards $3,000 to the "most human" chatterbot. The bots, which face off against humans, often use trickery, such as deliberate typos, to convince their interlocutors that they are the genuine humans. The prize also promises $25,000 to the creator of the first chatterbot that can convince judges that the human is the bot, and $100,000 for the first one who can beat a test that includes visual and auditory input as well as text.
These last two prizes have yet to be awarded, but even imperfect chatterbots have their commercial uses. Companies from IKEA to Lloyd's Banking Group have used "automated online assistants" to interact with their customers. Famously, the software agent Siri, a spin off from a Pentagon-funded "cognitive assistant" to help soldiers, launched on the iPhone 4S in October 2011.
Most people wouldn't confuse Siri or IKEA's chatterbots with actual humans, but not so for the spambots that send flirtatious chats to AIM or Yahoo! Messenger users, enticing them to follow dubious links. AIM spam has seen a resurgence this year, further vindicating Turing's predictions of at least some computer programs are successfully passing themselves off as humans.
Turing predicted that, once computers can pass the Turing test, "the use of words and general educated opinion will have altered so much that one will be able to speak of machines thinking without expecting to be contradicted."
In keeping with the reigning behaviorist ethos of 1950, Turing used the word "think" to refer not to internal mental states, but to measurable outward actions. Turing rejected the objection that a machine could not be said to think because it wouldn't feel like anything to be a machine, pointing out that, epistemically speaking, one has no way of knowing for certain that anyone other than oneself experiences feelings.
Turing also dismissed the religious argument that a machine could never be imbued with a soul. He wrote, "In attempting to construct such machines we should not be irreverently usurping His power of creating souls, any more than we are in the procreation of children: rather we are, in either case, instruments of His will providing mansions for the souls that He creates."
As artificial conversationalists become increasingly believable, Turing may be remembered as being among the first to have welcomed non-biological agents into the club of thinkers. If that happens, it would be all the more tragic that British government came to view the man as less than fully human.
Turing, who today is regarded as a war hero for deciphering communications created by the Nazi's Enigma machines, was also gay, and homosexuality in Britain at the time was punishable by either imprisonment or "chemical castration," that is, mandatory injections of synthetic female hormones meant to weaken the libido.
When Turing was arrested, in 1952, he chose the latter. He committed suicide two years later.
As for Dr. Epstein, who is also a computer programmer and actually directed one of the Loebner Prize competitions in the early 1990s, he continued looking for a partner online. As he explained on WNYC's science broadcast Radiolab, not long after parting ways with Ivana, he began unknowingly exchanging emails with another chatterbot, only to be notified by the bot's programmer, who recognized Epstein's name.
"There are hundreds of these things out there, maybe thousands," he said. "That's what's coming."