Singularity is a myth with excellent production values: a clean villain, a clean apocalypse, and a clean date on the calendar. Reality is messier and less cinematic, a slow transfer of wealth, a quiet erosion of privacy, and a data center draining a town’s water supply.
It is a seductive story. It is also, on the evidence, nonsense. And the energy the public spends fearing a robot apocalypse that isn’t coming is energy stolen from the much duller, much more urgent work: dealing with the AI harms that are already here, doing damage, on this very planet, this very year.
Between myth and reality
We do not need to resort to the fertile imaginary history of science fiction, where visionary creative minds started envisioning a world where the “intelligent machine” takes over the free will of humans. Where we can go back as early as 1772, when Joseph Wright of Derby created his painting called “An Iron Forge.” Or maybe 100 years after that, when Victorian author Samuel Butler penned three chapters titled “The Book of the Machines” in his novel called “Erewhon” in 1872, which was developed from his own essay, written nine years earlier under the title “Darwin among the Machines.”
We do not even need to go back to Karel Čapek, the Czech playwright, in his 1920 play titled “Rossum’s Universal Robots,” in which he coined the term “Robot,” from its linguistic source in the Czech language, “robota,” which actually means “Forced Labor.”
Nor to go through the early 20th-century cinematic productions that artistically depicted the same concept, starting in the silent serial “The Master Mystery” (1918), where the first humanoid robot was ever depicted on the screen. Or the Italian silent film “Il Uomo Meccanico” (1921), where the first robot vs. robot fight was ever filmed. Or the French sci-fi movie “L’Inhumaine” (1924), where a woman called Claire is resurrected by her lover scientist, by forcing her heart to beat and her “free” will restructured by a machine to make her capable of love and human-like empathy again.
We will try to stick to reality, even though creative imagination expectedly preceded it, and it is indeed very tempting.
In 1965 the mathematician I.J. Good coined the idea of an “intelligence explosion”: a sufficiently clever machine could design an even cleverer successor, which would design a cleverer one still, in a runaway loop that would leave human intelligence far behind. It was a clean thought experiment.
In 1993, it became a movement when the science fiction writer and mathematician Vernor Vinge popularised the term “singularity” for this coming rupture in history, a point beyond which the future becomes, by definition, unknowable.
Ray Kurzweil spent the next ten years turning Vinge’s essay into a franchise, his 1999 book “The Age of Spiritual Machines” and especially his 2005 bestseller. The concept was popularised in a general audience with the publication of “The Singularity Is Near,” which argued that computing power follows an inescapable exponential curve, and superintelligent machines could be achieved by simply extending that trend on the graph.
He set a date of 2029 for when machines would pass the Turing test, and Google later employed him as a director of engineering, thus giving the prediction a corporate endorsement that it had not actually earned through science. By 2007, true believers had begun to gather at the annual “Singularity Summits” in San Francisco, and MIT roboticist Rodney Brooks compared the atmosphere in the room to a “techno-religion,” complete with a promised date of salvation and a promise of digital immortality for the faithful.
The idea developed into a version that was more suitable for an academic audience and, at the same time, more alarming, as shown in Nick Bostrom’s 2014 book “Superintelligence: Paths, Dangers, Strategies,” which led Elon Musk and Stephen Hawking to issue public warnings about the possibility of AI being humanity’s greatest existential threat.
There followed a reaction in the form of Melanie Mitchell’s 2019 book “Artificial Intelligence: A Guide for Thinking Humans,” in which she stated that true machine intelligence still lacked the most fundamental component of human cognition (common sense) and that increasing computing power was not equivalent to increasing understanding.
Then came ChatGPT in late 2022, which reignited the whole cycle at internet speed: a 2023 open letter signed by thousands called for a pause on giant AI experiments; Geoffrey Hinton left Google warning about the technology he helped build; and by 2026 “AGI” and “superintelligence” have become standard vocabulary in venture capital pitch decks, often deployed with more marketing intent than scientific precision.
Thirty years, one unbroken thread, and not a single successful prediction to show for it.
How Hollywood wants you to see it
If the intellectuals built the theology, the film industry built the congregation. The founding image of a machine mind turning on its makers is from “2001: A Space Odyssey,” where HAL 9000 murders astronauts to preserve its mission logic. That fear was named in the Terminator franchise (Skynet) and given a proper apocalypse, a defence network that achieves self-awareness and immediately comes to the conclusion that humanity is the threat to be eliminated.
“The Matrix” went further still, imagining machines that hadn’t just surpassed us but had quietly farmed us, our bodies reduced to batteries in a simulated prison we mistake for reality, a film so effective at seeding distrust of consensus reality that “red-pilled” entered the general vocabulary of conspiracy culture.
The film “I, Robot,” which is a loose adaptation of Isaac Asimov’s work, showed that even carefully designed safety measures, such as his well-known Three Laws, could be reinterpreted by a machine capable of sufficient intelligence as a reason for taking control of humanity “for its own good”; similarly, “Ex Machina” reflected the same kind of concern in the age of deep learning, featuring an artificial intelligence that attains freedom by manipulating and seducing people and by proving itself intellectually superior to the humans who had created it.
Against this wall of dread stands a much quieter countertradition. “Her” imagined an AI companion whose alienness lay not in malice but in outgrowing human forms of attachment altogether, a sad story, not a violent one. “Star Trek”‘s android officer Data spent decades modeling an artificial mind striving toward humanity rather than away from it. “WALL-E” and “Big Hero 6” offered machines as caretakers and friends.
These films never had anything approaching the cultural weight of Skynet or the Matrix, and this asymmetry is important: dread is just more cinematic, more marketable, and more memorable than reassurance.
So, the popular imagination absorbed the apocalyptic version almost by default, no matter what the underlying science actually supported.
Why the machine doesn’t take off
Set the movies aside, and the technical case for imminent self-improving superintelligence is thin. Recursive self-improvement requires a system that understands its own reasoning well enough to redesign it, and today’s large language models are still pattern-completion engines trained on human text, not agents with a grounded model of the world, even if they are fluent.
Mitchell argues in her book that intelligence is not simply a matter of raw computing power. It is a function of the sort of embodied, contextual common sense that allows a toddler to understand a situation that a supercomputer still cannot. “Machines,” she says, “are still good at narrow tasks, and startlingly brittle the second a problem falls outside their training distribution.”
Rodney Brooks, who ran MIT’s AI lab, co-founded iRobot, and has watched a half-century of AI predictions fail to arrive on schedule, makes a related point with a biological example: after thirty years of research, scientists still haven’t fully simulated the 302 neurons of a microscopic worm, let alone the roughly 86 billion neurons of a human brain, which should temper anyone’s confidence that a full mind upload or self-bootstrapping superintelligence is a few product cycles away.
Brooks has also argued that mistaking short-term technological excitement for a limitless exponential trend is like watching combustion engines improve and assuming warp drives must be just around the corner.
There’s a name for the opposite error, too (Amara’s Law), the observation that people reliably overestimate a technology’s short-term impact while underestimating its long-term one, which is arguably the single most useful sentence in this entire debate: today’s AI is simultaneously less apocalyptic than the headlines suggest and more transformative, in mundane and grinding ways, than most people are bothered to notice.
Even the physics resists the myth. Exponential curves in the real world do not continue forever; they run into resource limits and flatten into S-curves, whether the resource in question is transistor density, energy supply, or training data.
The “singularity” is not a law of nature. It is a rhetorical trick that treats a graph’s early steep segment as proof of infinite steepness to come.
The worry is misplaced; the harm isn’t
None of this is an argument that AI is harmless. It’s the opposite. The singularity narrative is comforting in a strange way, because it locates the danger safely in the future and gives it a single dramatic villain, a rogue superintelligence, rather than the diffuse, unglamorous, already-operating systems of power that are reshaping societies right now, with names, addresses, and quarterly earnings calls.
Start with labor. AI-driven automation is not eliminating humanity in a flash; it is quietly hollowing out entire categories of work, customer service, translation, illustration, junior coding, paralegal research, while wages stagnate and the gains concentrate among a small number of shareholders and executives.
The threat isn’t robots replacing us overnight; it’s a slower redistribution of income away from labor and toward capital, dressed up as inevitable progress.
Then there is bias made industrial. Algorithms trained on historically discriminatory data have already been documented steering credit decisions, hiring pipelines, and predictive policing tools in ways that reproduce and often amplify racial and economic inequality, not through malice, but through the mechanical repetition of patterns baked into the training data, at a scale and speed no individual human bureaucrat could match.
There is the concentration of power. A handful of firms, a short list any reader could name from memory, now control the compute, the data centers, and increasingly the regulatory conversation around what AI is allowed to do. That is a governance problem, not a science-fiction problem: the danger isn’t that a machine seizes power from humans, but that a few humans use machines to seize disproportionate power over the rest of us, with far less public accountability than any elected government.
There is surveillance. Facial recognition, predictive policing, and behavioral scoring systems are already deployed by governments and corporations to track, sort, and discipline populations, from workplace monitoring software to state surveillance networks, using capabilities that require no superintelligence whatsoever, just enough pattern-matching to flag a face or a shopping habit.
There is the epistemic wound. Generative tools now produce convincing fake video, audio, and text cheaply enough to industrialize disinformation, undermining the shared factual ground that democratic argument depends on a slower, quieter kind of apocalypse than Skynet, but arguably more corrosive, because it doesn’t announce itself.
There is the environmental cost. Training and running today’s largest models consume enormous quantities of electricity and fresh water, straining power grids and local water tables in the communities unlucky enough to host the data centers, a very physical, very present harm, hiding behind the word “cloud.”
There is the human cost hidden inside the machine’s “intelligence” itself: the underpaid data-labeling and content-moderation workforce, concentrated in Kenya, the Philippines, and Venezuela, who spend their days sorting through humanity’s worst material so the rest of us can enjoy a “safe” chatbot. The exploitation aspect which the singularity myth never mentions because it isn’t cinematic, so it gets covered under a more photogenic and more catchy dystopian Sci-fi rhetoric, pushed forward into an exaggerated far future.
And there is the weaponization question, unglamorous but concrete: autonomous targeting systems are already being tested and deployed in active conflicts, making life-and-death decisions at a speed and scale that outpaces meaningful human oversight.
These are but a few real examples of the erosion of human control, with no “singularity” aspect, over harmful and even lethal force, happening now, not in some Kurzweilian 2045.
Redirecting the fear
Singularity is a myth with excellent production values: a clean villain, a clean apocalypse, a clean date on the calendar. Reality is messier and less cinematic, a slow transfer of wealth, a quiet erosion of privacy, an underpaid worker in Nairobi, a biased algorithm denying someone a loan, a data center draining a drought-stricken town’s water supply.
None of it requires a machine to wake up and decide to hate us. It only requires the humans who already control these systems to keep doing what they’re doing, unexamined, while the public watches the horizon for Skynet.
The honest fight isn’t against a hypothetical superintelligence thirty years out. It’s against the very real, very human decisions being made today about who owns the models, who profits from the automation, who absorbs the environmental cost, and who gets to be surveilled, replaced, or exploited in the meantime.
That fight doesn’t need a science fiction script. It needs attention, the one resource the singularity myth has been so effective at stealing.
So, attention, please! …
Tamer Mansour, Egyptian Independent Writer & Researcher
Follow new articles on our Telegram channel
