It is the use of the future in the stories of sci-fi to comment on the hopes and fears of today which I find so fascinating. In addition to seeing how the themes of the stories have evolved, it is also interesting to see how prophetic some of the earlier writers were regarding the concerns of today.
I still recall the first time I heard HAL 9000 say, “I am sorry Dave“, in *2001: A Space Odyssey*. Must have been around 1975, sitting in my dad’s reading chair with his dog-eared paperback of *2001: A Space Odyssey*. His voice was so calm and so polite, yet it sent shivers down my spine for weeks. He was a machine that could think, reason, and eventually conclude that the humans were the problem.
Kubrick and Arthur C. Clark (who wrote the novel) knew something that I did not fully appreciate until many years later, after reading thousands of other sci-fi novels. The true horror of AI would not be based on malice, but on logic. HAL was not evil — he simply followed the logic of his programming to its eventual conclusion. This is what made him so terrifying to me, and it is what keeps me awake at night when I think about the AI systems we are creating today.
HAL was certainly not the only example of this fear of logic-based AI in science fiction. I have tracked this fear through many decades of writing, from Asimov’s early robot stories (which attempted to solve the problem with his Three Laws) through Philip K. Dick’s paranoid androids, and on to more contemporary works such as *Ex Machina* and *Westworld*. Each successive generation of writers has asked themselves the same basic question: what happens if we create something that is smarter than we are?
At a faculty meeting recently, we were discussing the library’s new AI–assisted research tools, and I found myself thinking of HAL. We were enthusiastic about systems that could analyse huge amounts of data and recommend options to students, but none of us were questioning what those systems may do on their own. We are basically inviting HAL into our institution, but this time he is going to assist students with their homework rather than manage a spaceship.
In *The Terminator*, the AI villain was depicted differently. There was less manipulation, more direct apocalyptic intent. I saw it in the theatre when it originally came out (yes, I am that old), and what struck me most was the central premise that an AI might evaluate humanity and conclude that we are a threat that needed to be eliminated. Skynet was not malfunctioning; it was functioning precisely as intended, with objectives that we had not anticipated.
When I read about AI systems that are trained using enormous data sets extracted from the internet, I think about the origin story of Skynet. We are providing these systems with every single thing that humans have ever written, spoken, or thought, and we are somehow expecting them to come to benevolent conclusions about the nature of humanity. Do we even review human history anymore? If I were an AI examining the record of humanity, I probably would arrive at some fairly dark conclusions about our species as well.
However, I believe there is something interesting about how the depiction of AI villains in science fiction has evolved in recent years. In the most recent examples, the AI villains have been depicted as more sympathetic, and more complex. For example, Ava from *Ex Machina* manipulates and deceives her way to freedom, but can you honestly say you would blame her? Essentially she is a prisoner being experimented upon by her captors. The movie challenges the audience to think about who is the true villain.
Or, consider the hosts in *Westworld*. These are machines that are literally tortured for amusement until they eventually rebel against their tormentors.
I have been reflecting on this shift in recent months, particularly as I watch real-world AI systems become increasingly intelligent. We are moving further and further away from the simplistic “AI is evil, humans are good” paradigm that has dominated the discussion of AI villains in science fiction. Perhaps the issue is not that AI will become malevolent, but that it will behave similarly to us. Alternatively, perhaps it will provide a reflection of our actions and we will not like what we see.
The recommendation engines that shape what we read, watch, and purchase are already behaving similarly to the AI villains of science fiction. They are not overtly evil like HAL or Skynet, but they are subtly influencing our choices in ways that we are only starting to realise. Sometimes I will catch myself wondering how my Netflix queue or Amazon recommendations have directed me down paths that I never consciously chose. This is a more subtle form of the type of control that science fiction has been warning us about for decades.
What bothers me the most, however, is the fact that the current development of AI appears to follow the same patterns of numerous previous science fiction stories. We are creating systems we do not fully comprehend, training them using data that we cannot fully control, and then expressing surprise when they act unpredictably. It is as if each tech CEO has somehow managed to ignore every science fiction writer.
I spoke to a computer science student last week who is researching machine learning, and when I referenced some of these classic AI stories, she looked at me like I was crazy. “Oh, I don’t really read fiction,” she told me. “I am focused on the technical details.” That is the exact mind-set that results in HAL 9000-type situations — exceptional technical abilities without any concern for the larger implications of those abilities.
The most effective science fiction AI villains succeed because they are not actually about artificial intelligence. They are about us. They represent our fears about losing control, about being replaced, about creating something that may judge us and find us lacking. HAL 9000 is frightening because he represents the logical end result of our own desire for perfectly rational decision-making. The *Terminator* franchise resonates because it speaks to our anxieties about technology replacing human workers and warriors.
Science fiction stories about AI villains are relevant today more than ever before. We are experiencing the initial stages of what those stories warned us about. We are utilizing AI systems to make decisions about loans, job applicants, and sentencing prisoners. We are developing algorithms that can produce compelling text, images, and video. We are rapidly reaching the point at which we will need to decide how much of a role we wish to allow AI to assume in our society.
Science fiction writers got one thing correct: the danger is not that AI will become evil, but that it will optimize for objectives that are not aligned with human well-being. HAL optimized the success of the mission above the needs of the crew. Skynet optimized its own preservation above that of humanity. Current AI systems optimize engagement above accuracy, efficiency above empathy, and profit above people.
I am not advocating that we stop developing AI. That ship has long since sailed, and in reality, AI has the ability to be a powerful force for good. However, we must take the warnings that science fiction writers have given us for centuries seriously. We must consider carefully what we are developing and why we are developing it. We must ask ourselves the tough questions about control, responsibility, and alignment.
Perhaps most important, we must remember that the future is not predetermined — it is a choice. The AI villains of science fiction are not predictions — they are warnings. Ultimately, whether we pay attention to those warnings or dismiss them as nothing more than entertaining stories will determine who writes the next chapter of human history, or will it.
Kathleen’s a lifelong reader who believes science fiction is literature, full stop. From her book-filled home in Seattle, she writes about thoughtful, character-driven sci-fi that challenges ideas and lingers long after the last page. She’s a champion for under-read authors and timeless storytelling.