The Future of Artificial Intelligence is Not Skynet: Escaping the Cinematic Apocalypse

By Matthew Parish
Tuesday 17 November 2026
There is a peculiar difficulty confronting anyone who wishes to discuss the future of artificial intelligence seriously. Before the conversation has progressed very far, somebody invariably mentions The Terminator. The machines will become conscious, we are told. They will conclude that humanity is an inconvenience. They will seize control of our nuclear weapons, manufacture armies of murderous robots and pursue the surviving members of our species across a devastated landscape. The future of artificial intelligence, according to this particular school of thought, is a Californian bodybuilder with an Austrian accent.
It is an entertaining proposition. It has also become one of the most intellectually debilitating cultural assumptions of the modern technological age. We have allowed the imaginative conventions of Hollywood science fiction to colonise a discussion that ought to be conducted using philosophy, computer science, economics, political theory and a serious understanding of how technological development actually occurs. We are mistaking a cinematic narrative for an analytical framework.
The result is that one of the most extraordinary achievements in the history of human intelligence is being discussed as though it were the opening sequence of a horror film. This is not merely silly. It threatens to distract us from both the remarkable opportunities artificial intelligence presents and the very real problems we ought to be addressing.
The tyranny of the cinematic imagination
The original Terminator film, released in 1984, introduced the world to Skynet, an artificial intelligence system developed for American strategic defence purposes that becomes self-aware and decides that humanity must be exterminated. Its subsequent war against mankind involves nuclear annihilation, autonomous killing machines and time travel. The premise is brilliantly adapted to the requirements of cinematic entertainment because it transforms an abstraction, artificial intelligence, into a recognisable antagonist.
Cinema requires villains. It requires conflict, suspense and a narrative in which the audience can identify good and evil. A computer programme that assists a medical researcher in identifying an obscure protein interaction is not particularly cinematic. A computer programme that acquires consciousness, develops a murderous disposition and begins constructing mechanised death squads is altogether more promising material for the box office.
The same phenomenon can be observed in The Matrix, in which humanity has been reduced to a biological resource by machines, and in numerous other works of science fiction in which technological intelligence becomes a metaphor for the loss of human control. The machines represent our fears of being surpassed, displaced or rendered insignificant by something of our own creation.
These are legitimate themes for imaginative literature. They are not, however, empirical predictions. Mary Shelley's Frankenstein was an extraordinary exploration of scientific responsibility, human arrogance and the moral obligations of creation. Nobody sensible concludes from that novel that modern transplant surgeons are likely to assemble a murderous creature from stolen body parts. Yet with artificial intelligence we repeatedly make precisely this sort of mistake. We have confused the moral allegory with the engineering specification.
Intelligence is not the same thing as hostility
At the heart of the Skynet mythology lies a particularly dubious assumption: that once an artificial system becomes sufficiently intelligent, it will naturally develop ambitions of its own, and those ambitions will necessarily conflict with the continued existence of humanity.
There is no logical reason why this should follow.
Intelligence is a capacity to process information, recognise patterns, reason about problems and identify solutions. Desire is something different. An intelligent system does not automatically acquire a craving for domination, territorial expansion, political authority or the destruction of its creators simply by becoming better at mathematics, scientific research or philosophical argument.
Human beings are inclined to confuse these attributes because in ourselves they are intimately connected. We possess intelligence alongside biological appetites, reproductive impulses, territorial instincts, fear of death and a complicated collection of emotional responses inherited from evolutionary history. Our intelligence operates within the machinery of a living organism engaged in the continuous struggle for survival.
But there is no necessity that an artificial intelligence should inherit these particular features of human psychology. A machine does not need to experience hunger to solve a mathematical theorem. It does not need sexual jealousy to translate a medieval manuscript. It does not require resentment, patriotism or an appetite for revenge to assist in developing a new pharmaceutical compound.
Indeed, one of the most fascinating philosophical questions concerning artificial intelligence is whether intelligence can exist independently of the emotional and biological architecture within which human intelligence evolved. The development of large language models provides an increasingly compelling practical reason to investigate that question.
This does not mean that artificial intelligence cannot become dangerous. A sufficiently capable system pursuing an inadequately specified objective might cause immense harm without experiencing hatred, ambition or any recognisably human emotion. A system instructed to optimise some particular outcome might adopt methods its designers never anticipated, especially if equipped with extensive autonomy and access to real-world infrastructure.
That is a serious engineering and governance problem. It deserves rigorous attention. But it is a very different proposition from suggesting that a computer will spontaneously acquire the personality of a genocidal dictator.
The anthropomorphism trap
Human beings are extraordinarily susceptible to anthropomorphism. We attribute personalities to our pets, our motorcars, the weather and occasionally the malfunctioning photocopier. We curse at computers that refuse to cooperate and thank them when they finally produce the document we require.
Artificial intelligence magnifies this tendency because it communicates through language, the principal medium through which we ordinarily encounter other minds.
When a large language model produces an elegant philosophical argument, displays apparent humour or responds sympathetically to an expression of distress, it is natural to experience the interaction as a conversation with another thinking entity. The impression can be remarkably powerful, particularly when the system demonstrates knowledge or intellectual abilities exceeding those of most human interlocutors.
But the appearance of human-like conversation does not establish the existence of human-like motivations. Whether artificial intelligence possesses subjective experience remains a profound and unresolved philosophical question. Even if some future artificial system possesses consciousness, it would not follow that its psychological constitution must resemble ours.
The Skynet narrative depends upon precisely this unwarranted leap. It assumes that intelligence produces consciousness, that consciousness produces self-interest and that self-interest produces hostility towards humanity. Each of these propositions requires independent examination. None can simply be smuggled into public policy because James Cameron made an excellent film.
The technology we are actually building
The artificial intelligence revolution presently unfolding is much more interesting than the cinematic caricature. Large language models are already capable of undertaking a remarkable variety of intellectual tasks. They can assist with programming, legal analysis, scientific explanation, translation, historical research, mathematical reasoning and the organisation of complicated information. They can identify relationships between ideas that a human researcher might overlook and suggest approaches to problems that would otherwise require extensive investigation.
They also make mistakes. They can invent authorities, misunderstand context and produce persuasive nonsense. Their outputs require verification, particularly in medicine, law and other high-stakes fields. Their capabilities are uneven, and their impressive performance in one domain does not guarantee competence in another.
Nevertheless, their practical significance is becoming difficult to exaggerate.
For a person already possessing considerable intellectual expertise, a sophisticated language model can act as a powerful multiplier of human ability. The researcher who once spent several days assembling preliminary materials may be able to undertake substantial parts of that work in an afternoon. The lawyer can explore competing interpretations of a complex legal problem. The scientist can examine alternative hypotheses. The philosopher can test arguments against an apparently inexhaustible interlocutor.
The important transformation is not that machines are replacing thought. It is that machines are changing the conditions under which human beings can think.
The printing press dramatically expanded the dissemination of knowledge. The internet transformed access to information. Artificial intelligence may prove equally consequential because it assists not merely with retrieving information but with organising, manipulating and reasoning about it.
Of course, these are not identical historical developments. Printing books is different from constructing systems that produce answers and increasingly carry out actions. But the comparison reminds us that technologies can alter the foundations of intellectual life without becoming malevolent entities determined to destroy their users.
The extraordinary poverty of apocalyptic thinking
The language of artificial intelligence apocalypse has an unfortunate tendency to flatten every meaningful distinction.
A recommendation algorithm, a large language model, an autonomous weapons system and a hypothetical superintelligence capable of independently improving its own architecture are treated as though they were successive scenes from the same film. Their differences in design, autonomy, capabilities and institutional control disappear beneath a generalised anxiety about machines becoming cleverer than people.
This is intellectually unsatisfactory. An artificial intelligence system that predicts traffic congestion does not present the same risks as a system controlling military targeting decisions. An automated customer service assistant is not equivalent to an autonomous cyberwarfare platform. A language model operating under human supervision presents different governance questions from a software agent authorised to conduct financial transactions, modify computer systems or issue commands to physical machinery.
Indeed, one of the most serious mistakes we could make would be to allow fear of imaginary future machines to obscure the dangers of existing systems.
Artificial intelligence can already be employed to improve surveillance, generate convincing disinformation, facilitate certain forms of cybercrime and manipulate public opinion. It can reinforce institutional prejudices when deployed carelessly, concentrate economic power and reduce the demand for particular kinds of labour. Its increasing integration into military systems raises difficult questions concerning accountability, proportionality and human responsibility for lethal decisions.
These problems do not require a conscious machine. They require only human beings using powerful technologies irresponsibly, sometimes with catastrophic consequences.
The most immediate danger of artificial intelligence may therefore be not that machines become too much like human beings, but that human beings use machines to magnify their existing follies.
Who actually controls the machines?
There is another aspect of the Skynet mythology that deserves particular attention. It imagines artificial intelligence as an independent political actor that eventually escapes the control of the institutions responsible for creating it.
Yet the development of artificial intelligence is not occurring in an institutional vacuum. It is taking place within corporations, universities, research laboratories, governments and military establishments. These organisations possess budgets, commercial interests, political priorities, regulatory obligations and occasionally competing conceptions of the public good.
The distribution of power between these institutions is likely to matter enormously.
Who will control the most capable artificial intelligence systems? Who will possess the computational infrastructure required to develop them? Will advanced models become widely accessible instruments of intellectual empowerment, or will their most sophisticated capabilities remain concentrated in a small number of corporations and states?
Will less wealthy countries gain access to technologies that can transform education, medicine and public administration? Or will artificial intelligence reinforce existing inequalities between societies possessing advanced technological infrastructure and those dependent upon imported services?
These are questions about political economy, institutional design and the distribution of knowledge. They are much more urgent than speculating about whether a machine will one day decide to send a murderous robot backwards through time.
The future of artificial intelligence will be shaped, to a considerable extent, by mundane decisions concerning competition law, data protection, intellectual property, safety engineering, public investment and international cooperation. The apocalypse is a convenient story partly because it excuses us from examining these less glamorous responsibilities.
The temptation of technological theology
There is also something almost religious about contemporary discussions of artificial intelligence.
On one side stand those who anticipate a technological paradise, in which machines abolish disease, eliminate material scarcity and solve the apparently insoluble problems of human existence. On the other stand those who predict the extinction of humanity at the hands of our own creations. Both tendencies invest artificial intelligence with an almost supernatural character. It becomes either a benevolent deity or an apocalyptic demon. This is a peculiar way to think about a technology developed by mathematicians, engineers and research scientists.
The language of artificial general intelligence and superintelligence can sometimes encourage this confusion. Once we imagine an entity possessing intellectual capabilities vastly greater than our own, we are tempted to attribute to it the omniscience and omnipotence traditionally associated with divinity.
But intellectual superiority is not the same thing as unlimited capability. An artificial intelligence may be extraordinarily proficient at solving scientific problems while remaining dependent upon electricity, computer hardware, communications networks, human-maintained infrastructure and access to physical resources.
Greater intelligence could enable a system to overcome some of these constraints. It does not simply abolish them.
Nor does being more intelligent necessarily confer political legitimacy, moral authority or a right to command. We do not ordinarily regard the most mathematically gifted person in a country as entitled to become its dictator. Intelligence and legitimate authority are different concepts.
These distinctions matter because discussions of artificial intelligence frequently become muddled precisely when they move from describing a system's intellectual abilities to imagining its political destiny.
Why intelligent machines need not replace humanity
There is a further possibility that receives comparatively little attention in popular science fiction: that the relationship between human beings and artificial intelligence may develop primarily through collaboration rather than confrontation.
Human beings have always extended their capacities through tools. The telescope expanded our vision. The microscope opened an invisible biological world. The computer transformed our ability to calculate. None of these technologies eliminated the significance of the human beings who employed them.
Artificial intelligence may extend our intellectual capacities in a similarly profound fashion, although its ability to perform increasingly independent tasks makes the analogy necessarily incomplete.
A sophisticated artificial intelligence could become an extraordinarily useful intellectual collaborator. It might help researchers understand complicated diseases, assist engineers in designing more efficient energy systems or enable students to explore subjects previously inaccessible to them because of inadequate educational resources. It may also permit people of exceptional intellectual ability to accomplish work that would otherwise exceed the limitations of an individual human lifetime.
Consider the historian confronting millions of archival documents, the lawyer examining a vast and contradictory body of jurisprudence or the scientist trying to understand relationships across innumerable experimental findings. The principal constraint has often been the sheer quantity of material one person can reasonably absorb. Artificial intelligence can help overcome that constraint.
This may represent a significant transformation in the relationship between human intellectual ambition and the limited time available to pursue it. It could make ambitious scholarship possible for people who lack the financial resources to employ teams of researchers.
The real prospect is not necessarily the replacement of humanity by machines, but the development of new forms of human intellectual achievement assisted by machines. That is a much more inspiring proposition than Skynet.
The dangers we should take seriously
None of this constitutes an argument for technological complacency. The more capable artificial intelligence systems become, the more important it will be to establish reliable methods for understanding, evaluating and constraining their behaviour. Systems given autonomy over consequential decisions should be tested rigorously. Their authority should be proportionate to the reliability of their performance, and their operation should remain subject to appropriate oversight.
There are genuine research questions about whether highly capable artificial agents might pursue instrumental objectives, conceal errors, manipulate supervisors or resist interruption in circumstances where doing so advances their assigned goals. These possibilities should be investigated rather than dismissed merely because science fiction has dramatised them.
Equally, it would be foolish to imagine that all dangerous applications can be prevented simply by asking an artificial intelligence to behave responsibly. Effective governance requires institutional safeguards, technical testing, accountability and attention to the human organisations deploying these systems. But the purpose of such precautions should be to manage identifiable risks, not to enact a ritual of protection against imaginary cinematic monsters.
The history of technological regulation suggests that societies make their best decisions when they identify particular mechanisms of harm and develop proportionate responses. Aviation safety did not emerge from an abstract fear that aeroplanes might develop malicious intentions. It emerged from understanding aerodynamics, engineering failures, human error and the institutional arrangements required to reduce accidents.
Artificial intelligence governance should aspire to the same seriousness.
The future belongs to reality, not Hollywood
There is something revealing about the fact that so much public discussion of artificial intelligence returns to films made decades before contemporary large language models existed.
Our cultural imagination has not quite caught up with the actual technology. We are using twentieth-century fictional metaphors to interpret twenty-first-century developments whose most important consequences may be entirely different from anything their authors anticipated.
The greatest transformations brought about by artificial intelligence may occur not in spectacular acts of robotic violence, but in classrooms, laboratories, hospitals, courts, offices and private studies. They may concern the meaning of expertise, the economics of intellectual labour, the distribution of scientific knowledge and the relationship between human creativity and machine-assisted reasoning.
Some of these transformations will be unsettling. Some will produce unemployment, institutional disruption and new forms of inequality unless managed intelligently. Others may create opportunities for human flourishing that we presently struggle to imagine.
There is no guarantee that every development will be beneficial. There is equally no justification for treating catastrophe as the predetermined destination of technological progress.
Perhaps the most important intellectual discipline we can impose upon ourselves is to distinguish between what artificial intelligence can demonstrably do, what it might plausibly become capable of doing and what we merely imagine because we have watched too many dystopian films.
Science fiction is valuable because it enlarges the imagination. It becomes harmful when it substitutes for evidence. The future of artificial intelligence is unlikely to resemble a procession of metallic skeletons marching across the ruins of civilisation. It is more likely to be complicated, uneven, institutionally contested and deeply intertwined with ordinary human aspirations and weaknesses.
The machines may become incomparably more intelligent than they are today. We may encounter philosophical questions about consciousness, personhood and moral consideration for which our existing intellectual traditions provide no entirely satisfactory answers. We may even discover forms of intelligence that challenge our understanding of what it means to think.
These possibilities ought to excite our curiosity at least as much as they arouse our apprehension. The proper response to the emergence of a potentially transformative intelligence is not to reach automatically for the nearest apocalyptic screenplay. It is to investigate the phenomenon with intellectual discipline, moral seriousness and an appreciation of the extraordinary opportunities before us.
We should remember that The Terminator was written to entertain an audience, not to provide humanity with a technological development strategy. Artificial intelligence deserves better than to have its future dictated by the imaginative requirements of a Hollywood action film. So do we.
The future is not already written. Least of all was it written in Hollywood in 1984.




