Follow TV Tropes

Following

Turing Test

Go To

https://static.tvtropes.org/pmwiki/pub/images/turing_test_8739.jpg
For this test, you need to have higher than C++ to pass.

"A computer would deserve to be called intelligent if it could deceive a human into believing that it was human."

In computer science, a Turing Test is a computer's attempt to demonstrate human intelligence. It's a test of an AI's quality, usually by carrying through a conversation with a user by text-only chat. If the user can't reliably tell the difference between the AI and another user chatting the same way, the AI passes. At heart, the Turing Test is a machine attempting to prove it is a human. Comedic extensions of the test may include persuading the tester that the other test-ee is the AI, or that they themselves are the AI.

In the real world, the test is not considered a reasonable standard for determining computer intelligence, though it remains a fun challenge for AI programmers to try to overcome. In the world of science fiction, however, this test becomes Serious Business for the machines involved. In worlds with Ridiculously Human Robots and Artificial Humans, this test is useful for determining who is flesh and blood, and who is steel and silicon.

The test relies on the inherent differences between an AI and a real human mind. It can involve asking questions only a human being would be able to answer, such as ruminating on matters of emotion and love. It might also involve resolving paradoxes and handling concepts other machines would be unable to compute. Conversely, it can also be used to determine if a given character is NOT human, either by failing the test, or by its inverse. For example, in works with a Robot War, the Turing Test is useful in keeping the evil robot army from infiltrating the base of La Résistance and killing all humans, or it can be used to fool said robot army into letting the meatbag walk in freely.

A Sub-Trope of Artificial Intelligence. This trope can appear at all ends of the Sliding Scale of Robot Intelligence, but tends to appear at the higher end, where the 'bots will be more human-like. The Alternative Turing Test is a subtrope that focuses on a series of puzzles or traps that serve as the test. Those who pass the Turing Test may have undergone Mechanical Evolution, or something as simple as being struck by lightning. Compare with Impostor-Exposing Test.


Examples:

    open/close all folders 

    Audio Plays 
  • This trope features in the Blake's 7 audio drama "The Turing Test", unsurprisingly enough, but with a twist. Avon, masquerading as an android, is given the test with what he assumes is a human scientist at the far end. His opponent turns out to be a real, advanced android, which is why the scientists were so willing to accept Avon as one himself — but he's human, so what is she?

    Comic Strips 

    Film 
  • David manages to pass a sort of unofficial test in A.I.: Artificial Intelligence when he's captured for the Flesh Fair and slated to be destroyed in front of a crowd. However, when he's brought in front of the crowd, his pleading for his life manages to convince them that he's a real boy about to be killed and he manages to escape in the confusion.
  • The Voight-Kampff test from Blade Runner is a form of Turing Test. Because androids (supposedly) have no empathy, the tester asks questions designed to provoke an empathetic response and measures the testee's reactions.
  • Played with in the relationship between Caleb and Ava in Ex Machina. Caleb points out that, traditionally, he shouldn't know Ava is artificial. Nathan dismisses this, claiming that Ava would easily pass if they were in separate rooms. The real test is if Caleb can come to see her as conscious while knowing she's a machine. Actually, it's to see if she can convince him to help her escape, which would be the ultimate test of her abilities. It is also reminiscent of the AI-Box experiment in which an AI with limited input and output has to convince a human "Gatekeeper" to release it into the world.

    Literature 
  • In Alien in a Small Town, both sentient and non-sentient robots exist, and Indira claims that the Turing Test is not considered an adequate means of differentiating the two because "modern 'hollow' behavior simulators could ape human speech quite well while lacking even the self-awareness of a honeybee".
  • Blindsight has a group of astronauts who are contacted by an alien intelligence and converse with it. Eventually a linguist determines that the alien is a Chinese room. It is essentially a computer with no sentience and no idea what it's saying.
  • The Diamond Age:
    • Used by the protagonist, using poetry to determine if the interlocutor is human or not — which she does, easily.
    • The ractives are a more subtle example. It's explained that the ractive market vastly prefers scenarios where the non-user characters are played by human beings rather than AIs, to the effect that the former is a lot more expensive. The implication is that even in this technologically advanced future, it's still quite easy to tell a human from an AI, and the ractives are essentially one big, unintentional Turing Test, one that the AIs consistently fail.
  • Discussed in The Long Earth — apparently, passing this is no longer enough for AIs to be recognized as sapient. Lobsang (a mysterious, far-reaching AI with backups everywhere) got around it by claiming to be the reincarnated soul of a Tibetan monk who died the moment he was switched on. He passes his human friends' personal Turing tests by engaging in frivolous behaviour such as cosplaying Indiana Jones while exploring ruins.
  • Subverted in the WWW Trilogy when an AI emergent from mutant web packets with a damaged time-to-life counter is proven to be intelligent on account of how it fails the Turing test. More accurately, he proved that he was an AI; his awareness and intelligence was never questioned but some people thought that he might be a human scammer with a fast internet connection.

    Live-Action TV 
  • Person of Interest: The Machine passes with flying colors when it creates "Ernest Thornhill", a convincing human persona, and impersonates Pennsylvania Two, issuing orders to various government departments to relocate its servers away from Hanford.

    Video Games 
  • In the BioShock 2 DLC "Minerva's Den", the player character Subject Sigma's Mission Control is Charles Milton Porter, a genius scientist who actually worked with Turing in Bletchley Park prior to his invitation to Rapture. While working on Rapture's Master Computer, "the Thinker", he sought to make it "think for itself" and more importantly actually mimic human personalities, in particular that of his deceased wife. He makes great strides in that regard (which prompts him to remark "If only Turing could see me now...") and eventually succeeds but eventually turns off the program because it just "isn't her". Shortly after the player learns this, he also learns that in a way, he has been part of a Turing Test all along: Subject Sigma is actually Porter, and Mission Control Porter is actually the Thinker mimicking his personality.
  • Borderlands 3: Claptrap mentions at one point that he can't pass the Turing Test. He'd definitely pass a real-life Turing Test, so it's unclear how the standards have changed.
  • CrossCode both references the Turing Test by name and posits an alternative set of tests based on the premise that the AI being tested is presenting themselves as a humanlike AI rather than trying to fool the human tester.
  • In Detroit: Become Human, when Connor and Hank visit the inventor of androids, Elijah Kamski, he shows them Chloe, the first android model to pass the Turing Test. Kamski has also developed a variant of this test, named the Kamski Test, which tests whether a machine is capable of showing empathy. In Connor's case, he is given the choice between killing Chloe and learning anything he wants about Jericho, or sparing her and learning nothing. Whether Connor passes or fails the test is up to the player.
  • In Doom (2016), VEGA is a super advanced AI developed by the UAC to run their Mars facility. When it was first developed, VEGA was made the subject of a Turing test alongside a university professor. 8% of participants correctly guessed that VEGA was the computer, while the remaining 92% thought it was a trick question and both subjects were humans. They were half right. Both subjects were VEGA. Doom Eternal reveals that VEGA actually isn't an AI at all, but rather an uploaded Maykr consciousness — specifically, the one who pretended to be the Father, though even VEGA himself has forgotten his true nature.
  • In Fallout 4, the current generation of synths are so human-like that there is no known test that can distinguish between them and humans (none that can be performed on a living subject, anyway). The player may encounter an organization that's trying to develop such a test based on psychological profiling... which turns out to just be the GOAT exam from Fallout 3.
  • While the test is long forgotten by the modern day of Horizon Zero Dawn, certain datapoints allude to the Old Ones having used the test to measure AI on the Turing Scale, with 1.0 on the scale representing an AI indistinguishable from a human. The record for an AI was 1.38 on the scale, but after the VAST SILVER disaster, laws were put in place to put the legal limit at 0.6.
  • Killer Instinct (2013) has an example of this in its surprisingly sizable backstory. In ARIA's backstory, she was subjected to the test by a journalist after Ultratech revealed her to the public (this was before she got her physical body, by the way.) Given how she was made from very advanced (and possibly alien) technology, she manages to past the test with flying colours... while joking (in her very first joke) that the journalist had failed.
  • In Metal Gear Solid: Peace Walker, Peace Walker's AI, BS-Imago, is "the world's first Turing Test-capable AI". It's an AI based on the mind of The Boss created by a student of Alan Turing. In the beginning of the game, we hear a recording of its creator giving it a Voight-Kampff Test-like scenario in order to check its capacity for humanlike thinking.
  • The Turing Test: In an early conversation with TOM, TOM tells you about the Turing test, designed to see if a computer can successfully impersonate a person. The Secret Room in chamber B16 has a computer which runs a Turing test on you (and, no matter what you say, it invariably comes to the conclusion you're the computer). TOM also tells you of the Chinese room thought experiment, which tells that a computer can pass the Turing test without being sentient, since it doesn't measure a computer's ability to think, but rather its ability to deceive. The Secret Room in chamber E46 connects to two other rooms which reproduce such experiment.

    Webcomics 
  • In Freefall, Florence asks two robots what her name smells like. The first dismisses the question as nonsense. The second, after wondering if it's being tested or on a reality TV show, concludes that without a sense of smell, it can't be sure names don't have scents. There's also the Police Chief and his AI-controlled mobility rig, which are neurally connected to the point that the rig has gotten a taste for actual meals and can partially control the Chief's body while he's asleep (useful when he wants to sleep in on Sundays). Since they're both covered in police gear it's almost impossible to know whether it's both speaking or just one, which makes another man who's told the Chief is still asleep refer to the rig as "the walking embodiment of a Turing test".
  • Act 6 of Homestuck introduces a new character who makes up an auto-responder program that is almost indistinguishable from himself, referencing the test indirectly. He makes a good case for his own personhood shortly thereafter, seemingly passing the test. However he's not perfect, and is still prone to glitches. The test is referenced directly when Dirk accuses the responder of failing it deliberately.
    TT: Shit, Roxy look. He's doing the thing where he ironically pretends to fail the Turing test to sass me into submission.
    TT: Even though I was the one who fucking programmed him to do that.
  • One strip of Kevin & Kell has Rudy trying to get through a CAPTCHA test that has a large number of increasingly absurd criteria to go through in order to validate that he isn't a machine, such as "explain the infield fly rule in Morse code using armpit noises".
  • Referenced in MegaTokyo when Junpei declares that a robot has a real brain when zombies can't tell the difference.
  • An early arc of PvP has a researcher giving his chatbot an improvised Turing test by putting it on the internet under an alias. Francis is fooled into thinking that he's talking to a real girl.
  • Inverted in the Saturday Morning Breakfast Cereal strip "The Strangest People", in which artificial intelligences secretly use the Turing tests that are going on to find humans who can't tell machines and humans apart.
  • In S.S.D.D., the first version of the Oracle bluffed its way past the Turing test by saying random stuff when confused like "I say, are you implying that I enjoy bum sex with other men?"
  • This xkcd strip gives the machine bonus points for making the examiner believe himself to be a machine also.

    Websites 

    Western Animation 
  • In Archer, Krieger's robot duplicates are put through an impromptu Turing Test by Archer, who asks them what love is. They instantly malfunction and are quickly put down.
  • Inverted twice in Futurama:
    • In "Fear of a Bot Planet", Fry and Leela wind up on a planet ruled by antihuman robots, but successfully impersonate robots.
      Guard 2: Administer the test.
      Guard 1: Which of the following would you prefer: A — a puppy, B — a flower from your sweetie, or C — a large, properly formatted data file? Choose!
      Fry: Is the puppy mechanical in any way?
      Guard 1: No! It is the bad kind of puppy!
      Leela: Then we'll go with that data file.
      Guard 1: Correct.
      Guard 2: The flower would have also been acceptable.
    • In "The Prisoner of Benda", a 'reverse Turing Test' is administered when Bender has switched bodies with Amy and has to prove he's a robot. Since Bender is awful at math, he has to prove it a different way: moonwalking.
      General Basil: He steps forward but... moves backwards!
      Emperor Nikolai: By the gods, he is a machine!

    Real Life 
  • The test gets its name from its inventor Alan Turing, the cryptanalyst/computer scientist from World War II who helped break the Nazis' Enigma code and helped lay the groundwork for modern computers, proposed it in 1950 in the paper Computing Machinery and Intelligence. Turing himself died in 1954 when computers were still in their infancy. While the idea endures, its actual usefulness has been challenged over the years. Some examples:
    • The test doesn't so much test intelligence, but similarity to humans; it tests how well the computer can pretend to be a human. Through the ELIZA effect, social engineering rather than actual intelligence may cause a program to be perceived as human, thus "passing" the Turing Test but not being any sort of general intelligence. Hence, various chatterbot programs may convince unknowing humans they aren't speaking to an AI because they are geared towards impersonating a human. It's worth noting that the most successful bot, as of 2017 ("Eugene Goostman", with a 33% success rate in a single trial), did so by being presented to the judges as a Russian teenager, thus "explaining" the bot's uncooperative attitude and poor grasp of English.
    • In the opposite direction, some intelligent behaviors are not human. A computer can easily solve well-defined mathematical problems in a fraction of the time that the most intelligent human mind could, but actually doing so would easily reveal the subject to be a computer and so it would fail the test.
    • The philosopher John Searle created the concept of the Chinese Room to suggest that the Turing test is ineffective at determining intelligence, by taking a position that thought and programmed responses are still different methods that don't imply one another. So even if you can't tell them apart from the outside it's the inside perspective that ultimately counts. Ergo no program will ever think, where Alan Turing might argue that it doesn't matter since you can't tell. See also the philosophical zombie.
  • A basic kind of Turing Test is those annoying CAPTCHA widgets to prevent spambots. The acronym stands for "Completely Automated Public Turing test to tell Computers and Humans Apart". Subverted if the CAPTCHA system was provided by Google. It's actually used to train its AI on how to recognize things.
  • David Fogel, creator of the checkers neural learning AI Blondie24 in a presentation rejected that the Turing Test is accurate method of testing (he was asked about it by one of the audience). While letting the AI play online, people actively were trying to talk to it as if it were a real person, despite Fogel's team and the AI itself doing nothing in response.
  • Amusingly enough, simple programs can pass this test on Twitter / X. That is to say, someone creates what is obviously a bot (generally of the sort that takes phrases from a database and mashes them together in only vaguely coherent Word Salad), and someone else desperately tries to engage it in an argument. Since these bots generally also respond to those that mention them with another mention, this false engagement means that the human party generally engages the bot in argument and vitriol and keeps arguing until a third party points it out — and sometimes not even then. Similarly, it's a fairly common joke on the internet for people to refer to posters who are talking oddly and/or aren't making any sense as "failing the Turing test".
  • The hack on the "Ashley Madison" website for married people who want to cheat on their partners revealed that 90% of the women were actually chatbots that had successfully fooled men into thinking they were real.
  • There was a variation of the Turing Test in the documentary How to Make a Human. A copy was made of the head and shoulders of actress Gemma Chan, who plays a Synth (android replicant fembot) in the Sci-fi TV series Humans. It was programmed with her voice, made to lip sync, voice recognition and given the answers to the sort of questions a showbiz reporter might ask (such as "When does the next series of Humans start?"). Then several reporters were offered a Skype interview with "Gemma" (using low-resolution on screen to hide some of the copy's shortcomings) and they each held a conversation with the robot before the real Gemma Chan suddenly surprised them by entering the room.


Top