A.I. Experts Working on a Turing Test Replacement That Isn’t Just an Imitation Game
If we want to create truly sophisticated artificial intelligence, we’re going to have to set the bar a little higher than a test that Cleverbot can pass. That’s why AI experts are getting together this weekend to use their standard meatbag intelligence to come up with something a little more rigorous than just imitating human speech patterns.
At the annual Association for the Advancement of Artificial Intelligence in Austin, leaders in the field will get together to try to come up with a suitable replacement for the Turing test at a workshop called “Beyond the Turing Test.” The Turing test was introduced by Alan Turing, a computer scientist and mathematician who is portrayed by Benedict Cumberbatch as he cracks the Nazis’ Enigma code in The Imitation Game. Turing debuted the test in 1950, during which human beings communicate with both a computer and a person through chat software and vote on which is which. If human judges can’t reliably tell which is which, the artificial intelligence is said to have passed the test.
However, the test is very easy to manipulate, even though setting out to deliberately manipulate the results kind of defeats the purpose of engaging in it. For example, a bot recently passed by pretending to be a 13-year-old Ukrainian boy speaking to non-Ukrainian judges, which covered up anything strange about its conversations. While the bot technically passed the test, it didn’t prove an advancement in AI so much as how the test could stand to be improved.
There are already proposed alternatives such as the Ada Lovelace Exam that we’re pretty on board with, but we’ll have to wait and see what the experts come up with. Gary F. Marcus, a New York University research psychologist and co-chair of the workshop, told io9, “We have had an amazing response. All the top people in the field that we invited are coming to speak, and we have had a ton of interest and lots of media—which is very rare before an academic workshop.”
Marcus hopes the new AI benchmarks, whatever they may be, will help keep the public interested in computer intelligence and help drive the research forward to new, better, and hopefully not sociopathic things.