One of the main hurdles machines have to leap before they can successfully enslave humanity is being able to convincingly pretend they are one of us. Most science fiction assumes this will happen long after advanced robot bodies are available. The believability of an advanced humanoid robot that still acts weirdly inhuman is the basis of Arnold Schwarzenegger’s career.
But what if artificial intelligence became good enough to fool humans much earlier? What about now?
That’s what some claim happened Saturday in London, as an AI chat program became the alleged first to pass what is known as the Turing Test.
The Turing Test was created by computer scientist Alan Turing in 1950, in an attempt to determine whether machines can exhibit intelligence similar to or indistinguishable from a human. It’s a lot simpler and less dangerous than the Voight–Kampff empathy test from Blade Runner, but the tester has the same goal, separate the men from the machines.
As it did in London, the test often takes the form of a competition which pits live people and AIs against judges. For a program to pass the test 30 percent of the judges must be convinced the program they are chatting with is one of the real people to pass. Eugene Goostman scored 33 percent.
Eugene Goostman won the competition by employing a very dirty trick, but, after all, dirty tricks are a human trait so it shouldn’t count against him. The program told the judges it was a 13-year-old Ukranian boy, named, obviously, Eugene Goostman. That put some biases about age and possibly nationality in play, to a point where the judges were more likely to believe that Goostman might not be able to come up with convincing answers to their questions.
There was no requirement Eugene get the right answers, just that his answers sounded like they might be from a human.
The team that created Eugene, led by Russian Vladimir Veselov and Ukrainian Eugene Demchenko, took advantage of this loophole.
“Our main idea was that he can claim that he knows anything, but his age also makes it perfectly reasonable that he doesn’t know everything,” Veselov said.
“Going forward, we plan to make Eugene smarter and continue working on improving what we refer to as ‘conversation logic.’”
So, does it mean the end of humanity? Nope, Eugene is still pretty primitive, and uses some tricks that wouldn’t work in most situations to game the system.
At least one outlet is calling shenanigans on the claim entirely. Although the achievement was reported everywhere from tech blogs to The Washington Post, Techdirt takes issue with the techniques used by “Eugene” and the trustworthiness of organizer Kevin Warwick, and the validity of the Turing Test itself. It’s complaints are worth considering before arming yourself against a coming Cylon army.
[Source: The Verge]