Alan Turing was an English mathematician who was a key figure in code breaking the German “Enigma Machine” during WWII and is considered the father of the modern computer. He was also among the first to consider the possibility of artificial intelligence.
In 1950 Turing published a paper that opened with this line: “I propose to consider the question, ‘Can machines think?’” Stop and ponder for a second how groundbreaking of a question this was in 1950. Anywho, in his paper, Turing proposed two different “Imitation Games.”
In the first imitation game, there are three participants: (1) a man, (2) a woman, and (3) a judge. The three people are each in separate rooms and can only interact with each other via a keyboard and computer screen. The purpose of the game is for the judge to figure out which of the participants is the man. Under the rules of the game the woman will try to help the judge correctly determine which participant is which, while the man will try to trick the judge. Turing noted that “the best strategy for [the woman] is probably to give truthful answers. She can add such things as ‘I am the woman, don’t listen to him!’ to her answers, but it will avail nothing as the man can make similar remarks.”
The second imitation game is a modification of the first but where the man is replaced by a machine. The goal of this second game is for the judge to determine which participant is human and which is machine. A machine “passes” the Turing Test if the judge is unable to determine which participant is human and which is machine. In his 1950 paper, Turing stated:
I believe that in about fifty years’ time it will be possible to programme computers, with a storage capacity of about 10^9, to make them play the imitation game so well that an average interrogator will not have more than 70 percent chance of making the right identification after five minutes of questioning. … I believe that at the end of the century 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.
Thus, under Turing’s formulation, the test is “passed” if 30% or more of judges cannot correctly identify which participant is human.
It is interesting to think about what questions you’d ask to determine which participant in the imitation game is human and which is machine. Maybe asking very tough math questions would be a good route — ones that a human couldn’t possibly answer but a computer could. For example, asking whether 826,963 is a prime number (it is) wouldn’t be something a human could determine quickly, while maybe the machine would answer it. Of course, a truly thinking A.I. would know that in order to imitate a human it would need to refrain from answering.
An early A.I. theorist, Roger Schank, proposes in his dynamite book Tell Me a Story: Narrative and Intelligence, that a key aspect to human intelligence is our ability to store away an experience as a story and then retrieve it and tell it at the right time. Much of human interaction is telling each other stories, and a primary way we judge each other’s intelligence is the quality of stories we tell, including the stories we get back from others in response to our own stories. We intuitively know that “storytelling strongly reflects intelligence. Telling a good story at the right time is a hallmark of intelligence. One right time is when you are asked a question. Another right time is when someone says something to you and you respond with a relevant story.” More on storytelling and intelligence in this IFOD.
So, under Schank’s theory, for an A.I. to pass the Turing Test it would need to have a deep bank of experiences that have been converted into stories upon which to draw and be able to tell the right story at the proper time. In judging a Turing Test, maybe it would be a good strategy to relate your own experiences and see what the participants say in response. For example, suppose I said “I am so upset — I backed out of my driveway and into my neighbor’s car and will have to pay thousands of dollars to have it fixed.” A human response might be “Oh no! I did that once and it was right after we had moved into a new house and I hadn’t even met the neighbor I hit yet — what a way to meet a new neighbor!” Would an A.I. be able to fabricate such a story? It seems that it would need to be able to do so in order to pass the Turing Test.
The fantastic (but disturbing) movie ex machina concerns a robot A.I. and whether it can pass a form of the Turing Test. In the movie the participant, a young programmer, knows that the robot is an A.I. The question is whether he’ll develop feelings for her and conclude that she’s a thinking, feeling being that is worthy of being assisted in escaping her confinement. Great movie — I won’t spoil the answer by telling you what happened.
To date, no A.I. has successfully pass the Turing Test. In 2014 a computer program named “Eugene Goostman” allegedly passed the Turing Test when 33% of judges at the Royal Society in London thought it was human. But, there were only three judges and Eugene Goostman was imitating a 13-year Ukrainian boy so that some nonsense was potentially explained by poor English skills. A.I. experts generally don’t consider this to be a successful Turing Test.
How far away are we from having an A.I. pass the test? Some experts think that by 2030 we’ll have A.I. advanced enough to pass the test, while others think that 20 years is a more likely timeframe.
Is this your way of telling us that AI is the author of IFOD?
That’s very interesting! It’s also quite a coincidence that I started listening to an intriguing audiobook through Chirp entitled, “The Secret Life of the Mind: How Your Brain Thinks, Feels, & Decides”. The Turing Test is covered in one of the chapters. If you want to buy it, Chirp has it as a featured deal for 5 more days at only $2.99. I’ve never really gotten into audiobooks, but this one is changing my mind. I think a lot also has to do with whether I like the narrator’s voice.