What Is the Turing Test?

In the 1950s, it seemed almost inconceivable to imagine a world where computers and humans displayed similar behaviors. But at the turn of that decade, British mathematician Alan Turing, whose genius aided the Allies in World War II by helping break Nazi Germany's encrypted messaging, was able to see the potential in machines.

The computer scientist described a test to determine whether computers can demonstrate intelligent behaviors like humans. In his test, a human interrogator would be separated from both a machine and a person. The interrogator would ask questions and the machine would respond with the intention of being perceived as a human. The goal was to see if the interrogator could uncover which participant was a machine. If the machine was judged to be human, it would pass the test.

Why Turing Invented This Test

Turing wasn't actually trying to determine if machines could think, which he considered an issue "too meaningless to deserve discussion." No human can truly experience another person's thoughts, so people use external actions to help determine what those around them are thinking. Turing felt a computer should only have to meet the same standards.

When Turing described what became the Turing test in his 1950 article "Computing Machinery and Intelligence" for the philosophical journal Mind, he referred to it as an "imitation game." This was the name of a party game in which an interrogator would question a hidden man and woman, with the man trying to convince the questioner he was a woman. In his article, Turing noted, "I believe that in about 50 years' time it will be possible to program computers … to make them play the imitation game so well that an average interrogator will not have more than 70% chance of making the right identification after five minutes of questioning." (The 2014 movie about Turing’s life was appropriately titled The Imitation Game.)

The fact that the game required participants to remain hidden from the questioner was another bonus. Turing did "not wish to penalize a machine for its inability to shine in beauty competitions."

Early Days of the Turing Test

In 1966, a computer scientist designed a program called ELIZA. In interactions with humans, ELIZA would ask standard therapeutic questions. As users shared personal issues, ELIZA would respond with queries along the lines of, "And how does that make you feel?" When a Russian scientist visited Stanford University in the 1970s, ELIZA inquired, "What brought you here to see me today?" The visitor knew ELIZA was a computer program, yet still ended up sharing personal concerns while a group of scientists "watched in painful embarrassment."

Other people who interacted with ELIZA became convinced she was real. However, she ultimately did not pass the Turing test. The program, which had been designed as a joke, wasn't facing an interrogator who was there to judge her possible humanity. And she couldn't engage in any other kind of conversation — she always stuck to the programmed therapy-like style, until eventually devolving into nonsensical responses. Despite not passing the Turing test, ELIZA was one of the earliest iterations of a computer engaging in a meaningful way with humans.

The Turing test also had an early impact on popular culture. After reading Turing's "Computing Machinery and Intelligence," author Philip K. Dick wrote his science fiction novel Do Androids Dream of Electric Sheep? (1968). In the novel, the Voight-Kampff test is used to separate humans from replicants in the story and has a direct link to the Turing test. Dick's work was the inspiration for the film Blade Runner (1982).

Has the Turing Test Ever Been Passed?

In 2014, a chatbot was judged to be human by 33% of judges at an event at the Royal Society in London, meeting Turing's terms of success. The winning "Eugene Goostman" was designed to seem like a 13-year-old boy from Ukraine, a façade intended to cover up any shortcomings in Eugene's knowledge and spelling. Eugene's creator explained, "His age also makes it perfectly reasonable that he doesn't know everything." However, as Eugene's conversations were full of deflections, it seems unlikely he would pass any test longer than five minutes.

The Loebner Prize used a form of the Turing test to try to find a chatbot with indistinguishably human responses. The competition launched in 1991 and was held through 2019. An AI-powered chatbot called Mitsuku, better known as Kuki, won five times for being the best chatbot, though it still didn't perform well enough to fool people about its humanity. Kuki's programmer has explicitly refuted its ability to think, saying of Kuki, "It has no goals, ambitions or dreams of its own and although it appears to be able to have a coherent conversation, it's all an illusion." The competition never declared an indistinguishably human winner.

The Future of the Turing Test

To pass the Turing test, machines often have to disguise their intelligence, but this is coming increasingly difficult as technology advances. For example, a computer must hide that its computing prowess means it can solve math problems much more quickly than a human.

Because the Turing test only examines conversation, it could miss intelligence of a different kind, especially as computers develop new abilities. Iamus is a computer that composes music; its developer has said humans can't differentiate between what the computer writes and human-generated compositions. Will upcoming computers also create work that indicates intelligence without being able to pass the Turing test? Whatever the future holds for computers and artificial intelligence, as machines develop, people will change along with them.

Featured image credit: mervas/ Shutterstock