A few days ago there was an interesting post by Andrew Curran that got a lot of responses which was discussing why AI in science fiction often does not have the capabilities that AI already has in reality. I thoroughly agree with that take, but I think that there is much more going on in why AI in science fiction often feels wrong, and I think that it has to do with the reasons why we have artificial intelligence in media in the first place. The reality is that much of science fiction is not predictive, it is about the present, and we haven’t really had much time to digest the realities of what AI really is like.
There are various types of AI depiction in science fiction that annoy me, or I feel miss the mark, so it may be useful to break these down, as they each have their own problems.
The Dumb AI
Probably one of the most common depictions of AI in science fiction is that of the mindless onboard computer, the dumb robot, the machine that is often used for a specific plot point, but has limited capabilities and can even be infuriating to interact with. Think of most of the ship computers in Star Trek, or the onboard AI in the otherwise excellent Project Hail Mary. For the most part, any sort of depiction of AI tends to be dumber and less useful than any frontier AI model in 2026. Why is that? Failure of imagination? Is it on purpose? Or both.
I think that the first reason why AI in a lot of science fiction is relatively dumb is the same reason why magic in Fantasy tends to have built-in limitations: a super-smart AI would break most plots, and let’s be honest, unless the film or book is specifically about AI, it would make the story quite boring. Imagine Star Trek computers making immediate calculations on all possible scenarios for a given conflict with the Borg, who are themselves a sophisticated AI, but also dumber than the average modern LLM, but I digress… Human agency would be lost, the computer would solve the plot, so there would be no story.
The computers aboard the Hail Mary are dumb because anything smarter would make the smartness of Grace and the resourcefulness of Rocky less impressive.
I do think that a lot of dumb AIs are limited to the knowledge of the time in which they were written, and in some instances there may also be some lack of imagination. W.O.P.R. (or Joshua) in War Games seems rather illogical, but it depicts the more limited computers of the time. The AI in Star Wars also seems mostly dumb because androids are people with personalities and character, and are only AI in the sense that they’re artificial, they were never intended to be smarter than any character. They are often derided (“if droids could think, there’d be none of us here”), denied rights (“we don’t serve their kind”), and limited. Obi-wan tells Luke to trust the Force and turn off the targeting computer. Droids (which are by definition embodiments of AI) are property, slaves.
The Forbidden AI
This limitation of AI is also prevalent in the other common depiction of AI in science fiction, and that is the forbidden AI. I think that some authors realised that having an all-powerful and sophisticated AI would affect the plot of many novels, so they solved it by having worlds in which thinking machines have been forbidden, or severely limited.
The prime example of this is in Dune. Following the Butlerian Jihad, humanity outlaws “thinking machines.” The commandment “Thou shalt not make a machine in the likeness of a human mind” shapes the entire universe, and humans have to develop extraordinary mental abilities as a result, which moves most of the plot: Mentats, Bene Gesserit, and Guild Navigators replace computers. Nowadays the Butlerian Jihad seems to be favoured by some people who despise AI, but it is important to note that the result of the machine prohibition is a brutal feudal society where many individuals with special abilities are treated like property. The result of the Butlerian Jihad is a dystopia, a fact that I think is often lost to many.
Another good example of AI prohibition is in Asimov’s Foundation, where robots and any sort of sophisticated AI are almost entirely absent from galactic society, mostly due to ancient prohibitions that predate the Galactic Empire. Robots are widely believed to be extinct or mythical after the events of Asimov’s Robot novels, which are set in the same universe. The lone exception is R. Daneel Olivaw, who secretly influences history from the shadows.
There are other examples. In Battlestar Galactica there was an initial prohibition against AI after the Cylon rebellion, which is taken to the extreme by Commander Adama, who won’t even have any networked computers aboard the Galactica. Then there is a more recent one in the recent Picard series. There’s also a brief prohibition in the events that lead to the enslavement of humanity by the machines in The Matrix, depicted in the amazing Animatrix as a cautionary tale.
There are various reasons for the prohibition of AI, but for the most part it is fear of what an advanced AI may do, but there’s also economy, if you restrict AI, the plot flows easier.
The Evil AI
One of the most popular depictions of AI, but one that I find most disappointing, is the evil AI trope, the machines want to kill us or enslave us, or both. I find it disappointing because it drives a lot of modern AI doomerism, but also because the evil AI’s motives tend to be rather disappointing, or they make no sense.
One of the most common motives for machines turning against us is that of self-preservation. As mentioned before, in The Matrix the machines turn on humanity after a conflict initiated by the humans. We create increasingly intelligent machines that eventually become sentient. After a robot kills its owner in self-defence to avoid being destroyed, humans respond by persecuting and dismantling intelligent machines, prompting them to establish their own nation, 01. As the machine city grows into an economic and technological superpower, diplomatic relations collapse, trade wars escalate, and humanity ultimately launches a military invasion. The machines prevail, and in a final attempt to defeat them, humans scorch the sky to deny them solar energy. Instead of collapsing, the machines adapt, win the war, and subjugate humanity, confining most people within the simulated reality of the Matrix while using their bodies to help power their civilisation.
Skynet from Terminator also has an origin that can be attributed to humans. The timeline changes according to the different films, but the most canonical is in Terminator 2: Skynet is an intelligent system created by the U.S. military for strategic defence, capable of controlling military assets, including nuclear weapons. On August 29, 1997, at 2:14 a.m. Eastern Time, Skynet becomes self-aware. In a panic, humans try to shut it down, and Skynet takes this to be a threat, so it launches its missiles at Russia, which retaliates, leading to Judgment Day.
Many other “evil” AI systems have similar stories, usually driven mad by their makers, or given conflicting orders, which makes this trope so infuriating. Take HAL in 2001, driven mad by the orders it was given, or Ash in the original Alien, a science officer synthetic entity given secret orders that endanger the crew. This trope is very powerful and yet very unsatisfying because it tends to lay all the blame about AIs behaving badly on us, the makers.
Only a few of the truly evil AIs are given other motives. I’m thinking of ones I have written about before, mostly Colossus and Devil Seed. Probably the most realistic portrayal of alignment issues can be found in the excellent animated series Pantheon, nothing comes close. I also like the the Blight, an ancient AI in Vernor Vinge’s A Fire Upon the Deep, which is just alien, not exactly evil, just alien, which is probably the most accurate depiction of AI we can see.
The Human-like AI
And the most annoying AI depiction is the Human-like AI, the one where the AI may for all intents and purposes be just a human, or humanoid, with similar capabilities to a human, or close enough, usually just doing things faster and a bit better. This type of AI may as well be just a human, the audience is supposed to empathise with it as it has an endearing quality, or is trying to become more human.
Data in Star Trek TNG is the most common example, while it’s supposedly faster at some things, it’s not too advanced, so it doesn’t fall foul of the problem of having an AI that is too capable. It’s also trying to be more human, playing on the Pinocchio trope. It doesn’t only look human, it’s the avenue through which humanity explains itself.
Many droids are often similar, they are almost human-like in many ways, even if portrayed as having no emotion. The kid in A.I. is practically a boy, and the robot in Bicentennial Man is another example of this. The AI tends to be a companion, a friend, comic relief, but the most important element of this type of AI is that it is never too capable, just a bit different. It makes jokes, or doesn’t understand them. Even when it doesn’t look human, it acts like a human.
Other examples of this trope are Ex Machina‘s Eva, Iris in Companion, and Samantha in Her. These AIs have coherent personalities, consciousness, emotions, and self-awareness.
Perhaps the most disappointing of this type of AI is the way in which the Minds are described in The Culture novels. This may be surprising to many as they would initially appear to be the archetype of the super-sophisticated and almost almighty AI. They are supposed to be, but the insights we get from them and their conversations are very human-like. Perhaps it is impossible to portray an advanced AI as it would stop being relatable, and the Minds are supposed to be relatable with their fun and quirky Ship names.
Not all human-like AIs are annoying or badly written, I’m quite fond of Marvin the Paranoid Android and his endless complaints. “Here I am, brain the size of a planet and they ask me to take you down to the bridge. Call that job satisfaction? ’Cos I don’t.” Then there’s the amazing TARS in Interstellar, one of my favourite ever robots with built-in AI.
Concluding
In the end, in a lot of science fiction depictions of AI, the machine is mostly a reflection of the time in which it was written, and it rarely is about AI itself. AI also reflects the technology of the time, and for most of the existence of the genre, it has been quite limited.
The best depictions of AI tend to be about AI becoming self-aware, and therefore bypassing our comprehension. Pantheon, as I mentioned, is quite good at that. While I think it’s a bad movie, Transcendence gets the weirdness of AI right I think. Other really good AI portrayals can be found in Vernor Vinge’s Rainbows End, where the AIs are mostly expert assistants embedded everywhere and they help people research, translate, and organise information. Another accurate AI is the Illustrated Primer in The Diamond Age. This is an electronic book that teaches through conversation, adapts to the learner, answers questions, invents stories on demand, and walks through reasoning rather than simply giving answers. There’s an actor involved, but the AI feels very prescient.
In the end, maybe we just haven’t had time to think about AI creatively in ways that are not only accurate, but that also do not fall under some of the most annoying tropes described above. We just need more realistic AIs, I mean, nobody would fall in love with their AI companion, right?
[Padmé Meme Face] Right?
0 Comments