Artificial Humanity—The State of AI

by / 0 Comments / 275 View / June 19, 2014

Meet Eugene Goostman.

He’s a 13-year-old boy living in the Ukraine with a pet guinea pig and a father who is a gynecologist. He was born in Saint Petersburg, Russia in 2001. He’s equally charming and disarming; his sense of humor often makes you forget about his imperfect grammar. He travels around the world competing in knowledge competitions, many of which he has won. Oh, and he lives inside a computer.

Yes, Eugene Goostman is not a real boy. First developed in Princeton, New Jersey by Vladimir Veselov, Michael Gershkovich and Sergey Ulasen, Eugene’s brain and body exist inside of a laptop. He’s a chatterbot, or a machine that can sustain extended conversation.

He does not rely on a script, rather he uses natural language processing to understand your words by looking at context and observing your demeanor. He then consults his knowledge base and an algorithm to formulate a response, which manifests in a response through his interface. He was designed to imitate humanity, not necessarily to be the smartest cyborg in the room.

British computer scientist Alan Turing, largely considered the father of the field as well as of artificial intelligence, first mused in 1950 that by the year 2000, machines would be capable of fooling 30% of human judges into thinking they were human after five minutes of questioning.

From this prediction sprung the Turing test, which sought to examine just that—the extent to which artificial intelligence could trick humans.

Developed by PrincetonAI, a small group of computer programmers and technologists not associated with Princeton, Eugene became the first machine to pass the Turing test on June 7, 2014, by fooling 33% of its human judges into believing it was flesh and blood—and, depending on who you ask, this either marks a monumental turning point or a small blip on the radar of AI.

Many have questioned the validity and the scale of the achievement. Making Eugene a 13-year-old Ukranian boy means a few things—it means the programmers don’t have to answer for small errors in grammar, and his knowledge base can be limited and elementary and still fall within the parameters of the character. There’s another thing: Eugene has a big mouth. His use of humor and personality are argued to have distracted the judges from his lack of intelligence and non-human tendencies.

Anil Seth, professor of cognitive and computational neuroscience at the University of Sussex, said that claiming Eugene passed the Turing test is “a major overstatement which does grave disservice” to artificial intelligence.

Scientist Hugh Loebner claims the five-minute test is too short. The version he uses to conduct tests is 25 minutes. Others in the media have placed the University of Reading, the host of the 2014 Turing test, under fire for administering too simple of a test.

Rodney Brooks, MIT professor of robotics, made the distinction that the accomplishment “doesn’t make a damned bit of difference to robotics or AI.” In other words, he’s not without his critics.

The idea of an artificial machine speaking to humans is not entirely groundbreaking or shocking. Thanks to a generation of science fiction movies and novels, as well as our very own, pocket iterations of the idea (Apple’s Siri), it is safe to say that humans are becoming more comfortable and acclimated with the concept of making conversation, even connecting, with machines. And, ultimately, that may be what the critics of Goostman are missing.

Yes, he may not be the most intelligent or refined product of the AI wave, but he very well may be the most believable. And, in the end, that could make all the difference. John Denning, who has been working on Goostman’s knowledge base since his inception and remains the only native English speaker among the original development team, said, “We didn’t create human intelligence, it’s about convincing people, it’s about the human experience. If it’s believable, then cool, we really did accomplish something pretty darn significant.”

“Intelligence” and “thinking” are imprecise words, and no definition from neuroscience or psychology has been able to clearly define what they are for machines. But if the developers were able to fool actual humans into thinking Goostman was a real boy, then perhaps he is.

When it comes to artificial intelligence, perception is everything. And, for a generation that has grown up around technology and whose interaction with it has always been personal, unique and dynamic, Goostman could not have found a better audience or a better time to be alive. The more refined the system becomes, the more believable the sell, and the more the line between machine and human is, ultimately, blurred.

So, the question becomes, did Eugene Goostman really pass the Turing test? Well, yes, but that may be the wrong question. The right question looks something like this: does it matter?

The Turing test, ultimately, does not measure a machine’s inherent intelligence—which is why so many high-end AI companies aren’t looking to cater to the Turing test to pass it, for fear that it might actually slow down their progress.

The Turing test measures something quite different. It measures how believable the machines are and how human they are. While it may not be to artificial intelligence what the moon landing was to space exploration, it is perhaps more akin to the first time we sent a human in orbit. It is the confirmation of the idea, however ephemeral and distant, that we can accomplish something that defies our understanding of humanity. To that end, Eugene Goostman is one smart kid.

References

Anthony, Sebastian. “Eugene Goostman Becomes the First AI to Pass the Turing Test, Convincing Judges That He’s a 13-year-old Boy.” ExtremeTech. N.p., 9 June 2014. Web. .

Omidi, Maryam. “Man vs Machine: Turing Test Winner Hits Back at His Critics.” The Calvert Journal. N.p., 16 June 2014. Web. .

Shchetko, Nick. “Did ‘Eugene Goostman’ Pass the Turing Test? – Digits – WSJ.” Digits. The Wall Street Journal, 10 June 2014. Web. .

Ulanoff, Lance. “Turing Test Winner Eugene Goostman: The Inside Story.” Mashable. N.p., 12 June 2014. Web. .