Researchers find cunning loophole to help dumb AI pass the Turing Test

Alan Martin July 5, 2016

The holy grail of artificial intelligence is convincing a human that the AI is another human, and that they are therefore free to talk candidly about sports and current affairs. In 1950, Alan Turing developed the Turing Test, which involved a human and a computer being asked a series of written questions. If a third party believed the computer was really the human, then it would have successfully passed the Turing Test.

Researchers find cunning loophole to help dumb AI pass the Turing Test

That’s a very high bar to entry. In the past, the Turing Test has been troubled by a computer pretending to be a child, but according to a new paper from the University of Coventry, there’s an even simpler solution.

To paraphrase Ronan Keating, artificial intelligence says it best when it says nothing at all. turing_test_passed_with_silence

Here’s how it would work. The fifth amendment of the United States constitution permits people to decline answering questions if they believe they might incriminate themselves. If a machine says nothing, nobody can judge its thoughts.

Indeed, a sentient artificial intelligence may not fancy its chances in a Turing Test, and it could, therefore, be an intelligent decision to “take the fifth” and decline to answer any questions put to it. In its own way, it would have passed the Turing Test, as the judge would have no way of telling its motivations.

The trouble is that this is exactly the same response you’d get if you asked a radish or a slice of ham to take the Turing Test, only for entirely different reasons.

As co-author Kevin Warwick says: “This begs the question, what exactly does it mean to pass the Turing test? Turing introduced his imitation game as a replacement for the question ‘Can machines think?’ and the end conclusion of this is that if an entity passes the test then we have to regard it as a thinking entity.”

“However, if an entity can pass the test by remaining silent, this cannot be seen as an indication it is a thinking entity, otherwise objects such as stones or rocks, which clearly do not think, could pass the test. Therefore, we must conclude that ‘taking the Fifth’ fleshes out a serious flaw in the Turing Test.”

It’s a good point, albeit a slightly facetious one. The beauty of the argument (or what makes it infuriating depending on your perspective) is that it not only breaks the Turing Test, but also pretty much every alternative to it. The Winograd Schema challenge, which asks machines to identify ambiguous pronouns? Plead the fifth. The Marcus Test, that asks machines questions based on video clips? No comment.

Sometimes, keeping schtum really is in your best interests. A broken laptop may be smarter than we previously thought.

Images: Uhm Sapiens and Michael Cordedda used under Creative Commons