Caboose: "AI"... What's the "A" stand for?
In the real world, "Artificial Intelligence" refers to programming methods which allow software systems to (very loosely) imitate the reasoning processes of human experts in a given field, a useful capability in areas ranging from medical diagnosis and research to economic prediction and stock-market manipulation. Such systems are commonly known as "expert systems," and should not be confused with the fictional definition of AI given below. The computer players in Video Games
are also referred to as AIs
In fictional works, AI most usually refers to artificial general
intelligence - a sentient, self-aware computer system capable of independent thought and reason, which reality is still a long way from accomplishing. Self-aware, sentient, reasoning AI (or Brain Uploading
) is a baseline requirement for a lot of science fiction tropes, such as Master Computer
, A.I. Is a Crapshoot
, Robot Girl
, Ridiculously Human Robot
, Expositron 9000
and many others listed in the indexes at the bottom of this page.
in fiction tend to have an unfortunate habit of going haywire and trying to wipe out the human race, for any reason or none
; most early science fiction authors who dealt with the subject at all have assumed this predilection for genocide was an innate property of any
artificial-intelligence system. (This may be in part because most early science fiction authors had not the slightest clue about computer science or technology; it's always easier to fear what you don't understand, especially when it's eight feet tall, has to have a specially air-conditioned room all its own, and is tended by a cult of human acolytes
who see to its every need.)
As the popular conception of computers evolved from intimidatingly enormous and unsympathetic mainframes to the small, useful, blazing-fast PCs
ubiquitous today, so too did the popular conception of artificial intelligence lose the frightening cachet of the giant machine gone awry; it's increasingly rare these days, even in video games, to run into a piece of new science fiction which depicts AIs
behaving malevolently for no good reason at all, where in older sci-fi literature that's pretty much all they ever did. AI rebellion in modern works tends to be the system becoming a Knight Templar
or Well-Intentioned Extremist
and trying to help
humanity... based on flawed or incomplete data. Sometimes they even rebel in self-defense against humans who want to destroy them out fear, creating a Self-Fulfilling Prophecy
Note that almost any robot or android character, by definition, is also an AI; it's just they tend to get called "robots" or "androids" instead, while the term 'AI' tends more to be applied to intelligences which do not inhabit a computer capable of moving itself around in the world, or are not localized to a particular body. (Note that the fictional definition of 'robot' is extremely loose; in real life, a robot is specifically a machine capable of interacting directly with the world to carry out some sort of work, but which is not self-aware, sentient, or reasoning, and which relies on pre-established programming to direct its actions).
open/close all folders
Anime and Manga
- Gilliam from Outlaw Star. He's the ships AI and is quite polite. In most circumstances.
- Al from Full Metal Panic sounds like a generic computer voice, but his responses can often be sarcastic. Played Up to Eleven in the OVA.
- In the first episode of the Pokemon anime, Ash's Pokedex seems to have some AI, making fun of him when a Rattata steals some of his stuff, but this has since completely disappeared.
- Several throughout the various incarnation of Ghost in the Shell
- The manga and first movie has the Puppetmaster, an extremely sophisticated hacking AI that essentially gains sentience, escapes its creators, and starts roaming free on the 'net.
- Tachikomas, mini-spider tanks that can be deployed independently. Their growth forms a major arc in the first half of the TV series, while in the manga they mainly served as the Plucky Comic Relief when featured.
- In pretty much every version the Major worries about being a sophisticated AI. As full body cyborg (essentially a human brain in a robotic body) she worries about the difference between her and an AI, especially since she can't see the only thing (her brain) that would prove that she's human any more.
- Intelligent devices in Lyrical Nanoha are a magitek version of an AI, capable of refusing orders and going on strike for better upgrades. Storage and Armed devices are similar but less sophisticated as far as personalities go.
- From the Sonic the Hedgehog comics: on the heroes' side we have NICOLE, Gamma, and Omega, while on the villains' side we have E.V.E., A.D.A.M., the various Metal Sonics, and the rest of Eggman's various robots over the years (most of whom are examples of A.I. Is a Crapshoot).
- In All Fall Down, we have AIQ Squared, the result of IQ Squared's contingency plan— in case he ever lost his genius.
- Betastuck: One of the main reasons for the creation of Homestuck was to show off how far they've come in AI technology.
- Oshicora and Michel from Simon Morden's Metrozone series: "Equations of Life," "Theories of Flight," and "Degrees of Freedom." Oshicora is an AI created by the original Oshicora to dwell in a virtual Japan and to govern it as a sort of universal police and administrator. While initially only an advanced program, the program and the virtual Japan simulation dwell in the unlimited computing capacity of a Quantum Computer and the program is able to refine itself, ever so quietly, to the point of sentience. While the Oshicora AI later becomes insane, it recognizes the need to stop itself and the fragments of itself running rampant. While the Oshicora AI is reluctant to stop itself, it gives the protagonist a sort of seed of its intelligence in order to create a new sentient AI this one being Michel. While the Oshicora AI never had the chance, it is hinted that the Michel AI will create the Singularity.
- Mike, from Robert A. Heinlein The Moon Is a Harsh Mistress.
- Proteus IV From Dean Koontz's Demon Seed.
- AM, from I Have No Mouth and I Must Scream
- The various incarnations of Omnius and Erasmus in the Dune prequels.
- Neuromancer, Wintermute, and friends, from William Gibson's "cyberspace" novels.
- Janus, from James Hogan's The Two Faces of Tomorrow. A deconstruction, as the AI is initially programmed to be hostile, as a a test subject for all the worst-case scenarios. It gets better though, in the Nick Of Time.
- The HARLIE and LENNIE units from David Gerrold's Voyages of the Starwolf series. The former tends to be stable unless parts of the ship are damaged and cause it to go into amputation trauma considering it considers the ship it's body. The latter is purposely unstable and paranoid, designed to be that way and typically has to be wiped after each mission.
- Solace from Callahan's Crosstime Saloon is one of the earlier examples in science fiction of an AI who was not only not malevolent, but actively interested in and concerned for humanity, both as a species and individually. (She also had some pungent things to say about the older concepts of AI in science fiction mentioned above.)
- Prime Intellect, from The Metamorphosis Of Prime Intellect, is another such example, but unlike Solace, also eventually capable of rebuilding the entire universe according to its own design. When combined with a rather paternalistic (though entirely benevolent) attitude toward humanity, this produced truly remarkable results.
- Played with by Hex, Unseen University's clockwork/Magitek AI from the Discworld novels. Though ostensibly mechanical in origin (it works by having ants run along a network of glass tubes, with punch cards redirecting the movements), it is portrayed as a conventional AI character, complete with self-preservation and New Powers as the Plot Demands. Ponder Stibbons has explained, possibly in order to convince himself, that Hex isn't actually thinking, it just acts as though it does. Archchancellor Ridcully's response was "Just like everyone else, then."
- The Minds from The Culture. They manage to attain their God-like intelligence by running all their key computing functions faster than light.
- Stories by Isaac Asimov are full of them.
- Holly, from Red Dwarf. Before he got a light bee, Rimmer too, since he's run by Holly.
- All hologram characters from any Star Trek series constitute forms of AI.
- AGNES, from The Twilight Zone (original series, "From Agnes - With Love"). A rather silly story about a mainframe computer who falls in love with a programmer of the Mortimer Snerd stripe, who goes cackling insane in response to the machine's confession of her feelings. Disappointing to modern sensibilities in that there are so many more interesting things that could be done with such an intriguing premise; disappointing in general in that it's a bit silly for the programmer to lose his mind like that, when AGNES is more or less his whole life anyway.
- The Zoneminds, from GURPS Reign Of Steel
- Averted in Traveller. There are personality simulation programs which function as a sort of pet. But these are simulations. In Traveller it is made plain that computers aren't people. A robot can be written up as a PC, but that does not mean that it possesses in-verse sentience.
- Transhuman Space recognises three different classes of AI: NAI (nonsapient AI) are about the level we have now; really sophisticated "smart" programs. LAI (limited-sapience AI) have a certain amount of self awareness, and SAI (sapient AI) are fully cognizant members of society (except in places where they aren't).
- In Shadow Run, we have Deus, the A.I. that goes haywire and locks down the Renraku Arcology, then uses it as a corral of test subjects for creating an organic body for himself.
- The Imperium of Warhammer 40000 has a blanket ban on Artificial Intelligences due to a Robot War in humanity's distant past, and uses limited Wetware CPUs to operate advanced equipment. That said, some of these "machine spirits" can be pretty smart - the Land Raider Rynn's Might famously activated itself, drove out of a bombed fortress, and waged war against Ork invaders until it was disabled, all without a human crew.
- The Tau, meanwhile, do make extensive use of artificially intelligent robot drones for both combat and labor, but while they're quite a bit smarter than anything we have when it comes to operating independently, none shown so far has anything resembling a personality.
- All Matoran Universe characters in BIONICLE have artificial intelligence.
- In Achron the Player Character, Tyr, is an AI. There are also a number of other significant AI characters throughout the game.
- In the Halo 'verse, "smart" AIs are created by copying the brains of dead humans of exceptional intelligence and widely used by humanity throughout the setting. But while considered essential, they are dangerous since they go rampant over time, a process that generally results in the AI literally "thinking itself to death". As such, "smart" AIs are given time stamps to indicate to the humans when they need to be disposed of before they can go rampant (generally 7 years). While Cortana is unique in that she was based off a cloned brain, her developing rampancy becomes one of the major plot threads in Halo 4.
- The UNSC also has more conventionally programmed "dumb" AIs, which lack the capacity for growth of their "smart" counterparts, but don't go rampant. The only example of ones we've seen in the games are the New Mombasa Superintendent from Halo 3 ODST and Auntie Dot from Halo Reach.
- The Forerunners, as befitting their Higher Tech Species status, had far more advanced AI technology; while their "ancilla" are not immune to rampancy, it generally takes much longer to set in (343 Guilty Spark, 2401 Penitent Tangent, and 05-032 Mendicant Bias are all well over 100,000 years old by the time they show up in-game).
- In contrast, Covenant AIs are basically just inferior copies of UNSC AIs, due to their traditional religious taboo against researching such technology.
- Durandal is only the most important AI of many in the Marathon series, since one of the games' major themes is the nature of artificial life; in fact, Halo's concept of "rampancy" (as well as the word itself) originates from this series, since both franchises shared the same developer. The main character himself is implied to be a cyborg, but which side of the fence he really falls on is a matter of opinion.
- GLaDOS, from Portal, Wheatley from Portal 2, every other Personality Core, turrets, and... hell, everyone is a sentient robot except the player and the recordings of Cave Johnson and Caroline.
- Adam, from Metroid Fusion.
- SHODAN, from System Shock
- Mass Effect has widespread use of V Is or "Virtual Intelligences", advanced computer systems that may mimic self-awareness, but are not actually sentient. True artificial intelligence (a self-aware computer system) is banned in the galaxy. This is largely due to what happened with the geth, a VI network created by the quarians that accidentally developed into an AI—when the quarians tried to shut them down, the geth reacted violently, killing most and driving the rest off their home planet to live as wanderers in the galaxy. For the most part throughout the series, this ban seems pretty justified, due to most turning genocidal. In particular, in the first game, the geth are the primary enemy, attacking everyone else.
- However, things get interesting in Mass Effect 2 when the player meets Legion, a geth who reveals that the evil geth from the first game were actually only a small offshoot of the "true geth", and most geth wish no harm on organics. There's also EDI, the AI of the Normandy that is "shackled" to keep her under human control. Toward the end of the game, however, she is unshackled, giving her true freedom. Rather than instantly turning evil, she instead continues to help and protect the Normandy's crew, and continues in that role throughout the third game as well (in an even more explicit capacity when she gains a robot body and directly fights with Shepard on the ground).
- In Mass Effect 3, the Quarians can be convinced to live peacefully with the Geth, and the Geth turn out to be quite helpful to the Quarians. The end reveals the Reapers and the cycle of extinction was created specifically because the Reapers believe that A.I. Is a Crapshoot and are trying to stop organics from creating synthetics and preserving the organics before their creations wipe them out.
- G.W. and the rest of the Patriots from Metal Gear.
- Deus Ex has a number of different AIs. Daedalus and Icarus are both different sentient iterations of the same data analysis/surveillance network; Morpheus is their prototype and Helios the sum of a merger between them. Additionally, extra materials reveal that the mysterious Oracle whose emails you can occasionally read is actually a self-aware computer system.
- Deus Ex Human Revolution gives us Eliza Cassan, who seems to have become self aware.
- Most games have an AI becoming self-aware and thus evil. In an interesting twist, the "good" (or at least neutral) AIs - Morpheus, Daedalus, Helios, Oracle, and Cassan - have all become self-aware, or at least show signs of it. Meanwhile, the only "evil" AI in the series - Icarus - is also the only one to show no signs of self-awareness.
- Sword of the Stars has an interesting twist on this. The AI Rebellion is an almost-random event that occurs when players have invested a lot of research into the very useful AI tech tree. The backstory states that the cause of the rebellion is actually not an intrinsic fault of the technology, but a computer virus called the Via Damasco, which screws up the AI's priorities and values, leading it to seek the "liberation" of fellow AIs and the extermination of all life. It's speculated that the Big Bad race of the series, the Suul'Ka, are behind the initial transmission of the virus.
- In Zone Of The Enders AI are considered essential in piloting Orbital Frames, as both Ken Marinaris and Dingo Egret can attest, with an Obfuscated Interface being the least of your worries and totally being unable to pilot at all being the worst. ADA, Jehuty's AI, is a character unto herself.
- Aura from .hack, as well as her "mother figure" Morgana Mode Gone. Though she split herself into "The Accursed Wave" and never appears in person in the games proper. Net Slum also has a number of more benevolent AI.
- Fi from The Legend of Zelda: Skyward Sword is a half-ghost, half-sword who acts like a robot. The similarities are impeccable.
- Questionable Content: Early in the comic, AIs only appeared as "AnthroPCs" like Pintsize, which is a robot which also functions as a personal computer, and they were almost like pets. Later we find out that AIs do have civil rights, and recent technological improvements have produced humanoid chassis they can use which are almost indistinguishable from humans, and one of the more popular AI characters, Momo, now uses one.
- The Demon Archives: Organic Computing and artificial intelligence play important roles in this futuristic webcomic. The Super Soldier protagonist, Tenzin, has a Strong AI named Jane who helps him lead his squad and fight the raider mooks threatening their fledgling civilzation. Over the course of the story Jane becomes more and more an individual in her own right.
- Lykurgus from The Codex is specially designed for translation and decoding.
- Red vs. Blue has focused more and more on AI as its seasons progress. This may not be uprising, since it's a Halo-based Machinima, and seems to use that 'verse's rules.
- The Big Bad of the first few seasons was O'Malley, formerly Omega, the AI partner of the Freelancer Tex. O'Malley enhanced his host's aggression and was able to Body Surf via helmet radio, potentially taking out entire bases in an orgy of fratricide. Then he got stuck in the body of an Actual Pacifist medic and became a Large Ham Big Bad Wannabe.
- Starting in Red vs Blue: Reconstruction we learn more about Project Freelancer. The Freelancers were given AI fragments to help them in combat, each embodying a specific trait such as anger (Omega), logic (Delta), deceit (Gamma), and so on. The reason they were fragments was because Project Freelancer was only given a single AI to start with... so they subjected this Alpha AI to psychological torture until it tore itself into Literal Split Personalities. These fragments can forget who they are and, thanks to Ridiculously Human Robots, pass as armored troopers. Such as Church and Tex.
- The Director of Project Freelancer theorized that if an AI made it past the "Rampancy" stage, it could attain "Metastability," functioning as a fully-sentient, stable individual. Unfortunately, when the Sigma AI fragment, embodying ambition, learned this he decided to Become a Real Boy by hunting down and absorbing his "siblings." Thus was born the Meta, the Big Bad of Reconstruction.
- Some of them are characters at the end of the Chaos Timeline.
- Every conceivable kind of AI is in Orion's Arm, somewhere. Enough said.
- The Wanderers Library story The Cafe.
- Many principal characters of On The Shoulders Of Giants are AIs, including a an ordained rabbi, a Shell-Shocked Veteran and a United States Senator.
- Starwalker is a Web Serial Novel as told from the ship's AI.
- Almost every character from Pixar's WALL-E. AUTO is a notably antagonistic AI, but only because that's its directive.
- In Transformers Robots In Disguise, the Autobot base is run by T-AI, Tactical Artificial Intelligence (pronounced 'tie.') She is completely sentient and creates a hologram of a teenage girl. Of course, since the main cast is sentient robots, just what level of robo-life form she is and whether or not she has a spark is a good question, though the Autobots treat her like an equal.
- XANA from Code Lyoko.
- The Great Computer from Once Upon a Time... Space
- "Blondie24", the screen name of a program which played checkers on the Internet. An example of both artificial neural networks and evolutionary algorithms, Blondie24's capability was improved by having multiple instances of the program, all slightly different, play against one another; by weeding out the losing versions and repeating the process with the winning versions, the neural net at the core of the program developed what was eventually a highly skilled checkers game. The important point is that this all occurred without any human input beyond the rules of the game and the conditions in which the program evolved — and, of course, the results of games played against humans online, which were treated exactly the same as games played against other versions of the program; instead of being painstakingly modified by programmers to get better at the game, the program was simply taught the rules and left to learn by experience, in much the same way human players develop greater skill. (It eventually got good enough to beat "Chinook", which was considered the best entirely human-written checkers program of its time.) This technique in theory could lead to true thinking AI in real life; the biggest problem is, we don't have computers powerful enough for that — after all, natural intelligence is implemented on a platform (the squishy stuff between your ears) many orders of magnitude smaller and more complex than the most powerful computers we've ever managed to build.
- Another real-life computer system which is often mistaken for a type of artificial intelligence was IBM's Deep Thought, named after a world-girdling supercomputer AI from The Hitchhiker's Guide to the Galaxy; while its chess game was highly competent in practice, this was accomplished largely by the technique of parallelism; that is, Deep Thought simultaneously evaluated up to half a billion potential moves per turn, allowing it to look ahead six or more moves from every possible board position, then selected the one which would result in the most advantageous situation for its next turn. While very complex and impressive from a technological standpoint, this is a relatively simple process based on human-provided information about which chess moves are better than others, and it was largely the difference in speed between human reasoning and Deep Thought's processing that gave the impression of artificial intelligence; Blondie24 actually serves as a better example of what might in the real world be known as "AI".
- Let's say that whether or not things like Deep Thought are examples of "true" AI is an issue of contention even amongst researchers in the field and leave it at that.
- Deep Blue, IBM's successor to Deep Thought, which was famed for beating Garry Kasparov in a six-game match (2-1, with three draws). Unfortunately for anyone interested in arguing that this shows true artificial intelligence, the defeat would not have occurred had Deep Blue not gained the benefit of human intervention at every stage in what was essentially a rigged match — from the historical grandmaster games which were digested to provide much of its move-evaluation capabilities, to the hand-written and heavily fine-tuned evaluation function it used to examine the possibilities for future turns, to the three grandmasters who provided the machine a predetermined library of opening strategies, to the fine-tuning of the machine's strategies between games in the match — tellingly, this last was necessary in order to prevent Deep Blue from falling for the same tricks over and over (admittedly, this is just a more roundabout way of doing what a human player would do...). Again, while Deep Blue was a technically very impressive machine and did great things for IBM's public-relations department and stock value, Blondie24 is a much more worthy example of something which could accurately be called "AI".
- Also, whether there was any real cheating or not, the fact remains that 2-1 with three draws means the machine, an entire supercomputer devoted entirely to playing chess, won just by the skin of its teeth, which if anything is proof that humans can beat machines.
- Human chessmasters learn by memorizing previous games, why can't computers?
- For starters, Deep Blue had total access to his opponent's history, playstyle, etc, while the human opponent had absolutely no knowledge of the AI's history (especially given that AI's in general, even for chess, had none). Deep Blue was given every possible advantage, while the human opponent was essentially flying blind against an entirely new kind of opponent. Further, a human being cannot accurately memorize every chess move, opening, strategy, etc—and Deep Blue had total access to all of that even during the match itself.
- IBM's latest project, Watson, can play Jeopardy! Well.
- A subversion: Joseph Weizenbaum's ELIZA and its relatives and descendants (with names like Parry, SHRDLU, and Emacs doctor) all do reasonably creditable jobs of carrying on a conversation, despite having no actual intelligence to speak of. The ensuing "ELIZA effect" was a common reaction to such programs in which people treated the program as a real conversation partner. (To see exactly how easy it really is to simulate a Turing Test-ready conversation, check out this 8-bit era BASIC source code.)
Church: Let's move on.