In his December 2025 essay, "AI Is a Crock," published at Straight Line Logic, Robert Gore delivered a devastating attack upon the artificial intelligence industry and the extraordinary claims made on its behalf. Gore argues that the AI revolution rests upon a fundamental misunderstanding of human intelligence, that its economic promises are exaggerated, and that its greatest political attraction lies not in the liberation of humanity but in the expansion of surveillance and control. His critique is deliberately sweeping, and some of his claims go further than the available evidence warrants. Nevertheless, his central warning deserves serious attention. Artificial intelligence need not become genuinely intelligent, let alone superintelligent, to become an extraordinarily powerful instrument of political domination.
The modern AI industry has developed an almost religious atmosphere around the prospect of artificial general intelligence and, ultimately, superintelligence. Vast sums are being invested in data centres, advanced computer chips, electrical infrastructure and increasingly elaborate machine-learning systems. Technology executives speak confidently about machines surpassing human capabilities across virtually every intellectual domain. Governments, meanwhile, compete to secure strategic advantages in what is presented as the defining technological race of the century. The implication is that humanity stands at the threshold of a transformation comparable to, or greater than, the Industrial Revolution.
Gore's response is to question the assumptions underlying the entire enterprise. What exactly is intelligence, and why should the manipulation of symbols and statistical relationships be regarded as equivalent to human thought? Contemporary large language models are trained on immense quantities of text and other data, learning complex patterns that enable them to generate plausible responses to questions. Their achievements can be remarkable, but does the ability to produce sophisticated language establish that a machine understands what it is saying?
Gore argues that the developers of AI have approached the problem backwards. Rather than beginning with a satisfactory understanding of human intelligence and attempting to reproduce its essential characteristics, they have constructed computational systems and then treated their operations as models of intelligence itself. He compares this reasoning to expecting a clock to reveal the fundamental nature of time. A clock measures temporal intervals, but its mechanism does not explain what time ultimately is. Likewise, a machine that produces intelligent-looking answers does not necessarily reveal the nature of thought, understanding or consciousness.
There is considerable philosophical force in this objection, although it should not be overstated. Engineers need not understand every aspect of a natural phenomenon before developing technologies that reproduce some of its effects. Aircraft fly without reproducing the biological mechanisms of birds, and computers perform calculations without employing the mental processes of human mathematicians. AI systems may therefore possess substantial practical capabilities without operating like human minds. The real error arises when successful performance is treated as proof that the underlying phenomenon has been explained.
This distinction is particularly important in discussions of consciousness. Human intelligence is not merely the production of correct answers. It involves experience, intention, curiosity, judgment, practical engagement with the world and the capacity to pursue purposes that matter to the individual. Whether these characteristics can be reproduced artificially remains an unresolved philosophical and scientific question. The ability of a machine to simulate conversation does not settle it.
Gore places particular emphasis upon curiosity. Human beings and animals do not simply respond to questions supplied by others. They investigate their surroundings, identify problems, explore possibilities and develop interests that direct their behaviour. A child asks why the sky is blue or why an object falls. A scientist notices an anomaly and begins an investigation. Intellectual discovery frequently begins not with an answer but with the recognition that something requires explanation.
Current AI systems can certainly generate questions, propose research programmes and identify apparent gaps in existing knowledge. Some systems can also undertake extended investigations with limited human supervision. It would therefore be inaccurate to claim that artificial intelligence is incapable of formulating questions in any operational sense. But the deeper issue concerns the origin and significance of its purposes. Does the machine possess curiosity as an intrinsic motivation, or is it executing objectives established through training, programming and external direction? The distinction remains philosophically important even when the observable behaviour appears similar.
Gore's criticism becomes more politically significant when he considers the relationship between AI and consensus. Large language models are trained upon existing human cultural products, including books, academic papers, journalism, websites and institutional documents. Their responses are shaped by this accumulated material, together with subsequent training designed to encourage particular forms of behaviour. There is therefore an obvious danger that AI systems will reproduce prevailing intellectual assumptions, including the errors and prejudices of influential institutions.
The history of science provides abundant warnings against treating consensus as an infallible guide to truth. Scientific revolutions frequently involve challenges to established assumptions, sometimes by individuals whose views initially appear eccentric or implausible. A system trained to favour conventional answers may discourage precisely the forms of intellectual independence required for radical innovation.
Yet the claim that AI merely averages existing opinions is also inadequate. Modern systems can combine ideas in unexpected ways, identify inconsistencies, develop mathematical arguments and suggest unconventional hypotheses. They are not simply mechanical polling devices. Their limitations are more subtle. They may generate genuinely useful ideas while remaining vulnerable to systematic errors, misleading confidence, fashionable assumptions and biases introduced during training and evaluation.
The danger is therefore not that AI can never produce an unconventional thought. It is that millions of people may increasingly rely upon a relatively small number of systems to determine which thoughts deserve consideration. When intellectual inquiry becomes mediated through centralised technological platforms, the assumptions embedded in those platforms acquire enormous social influence.
This possibility should concern anyone who values intellectual independence. A society in which citizens routinely consult automated systems before forming opinions may gradually lose the habits of independent investigation. The problem is not simply that an AI response might be wrong. Human teachers, journalists, academics and experts are frequently wrong. The distinctive danger lies in the scale, speed and apparent authority with which automated answers can be delivered.
An error repeated by one journalist may influence thousands of readers. An error embedded in a widely used AI system may be reproduced across millions of conversations, business decisions, educational exercises and administrative procedures. If users cease to verify claims because the responses are fluent and apparently authoritative, intellectual dependence can become self-reinforcing.
Gore also raises the problem of model collapse, in which the increasing use of synthetic training data can degrade the quality and diversity of subsequent models. The danger is real, although research suggests that the consequences depend heavily upon how synthetic data are generated, selected and combined with authentic material. There is no universal law requiring every model trained partly on AI-generated content to deteriorate. Nevertheless, the broader concern is legitimate. An information environment increasingly populated by machine-generated material may become less diverse, less reliable and more difficult to distinguish from authentic human experience.
This leads to the phenomenon sometimes described as AI-generated work slop. Businesses adopt generative systems expecting dramatic improvements in productivity, only to discover that employees must spend substantial time checking, correcting and rewriting the resulting material. Work is not necessarily eliminated; it is redistributed. The apparent efficiency of generating a document in seconds can disappear when another worker must devote hours to verifying its factual accuracy, legal implications or practical usefulness.
None of this means that AI has no economic value. In software development, translation, research assistance, document analysis and many routine administrative tasks, the technology can deliver substantial benefits. The question is whether those benefits justify the enormous expenditure now being committed to infrastructure and model development. There is a considerable difference between demonstrating that a technology is useful and demonstrating that every company investing billions in it will earn an adequate financial return.
Gore's economic scepticism deserves attention for precisely this reason. The AI industry is characterised by extraordinary capital expenditure, intense competition and expectations of future revenues that remain uncertain. Data centres require expensive equipment, enormous quantities of electricity, cooling systems and continuing technological upgrades. The commercial life of specialised hardware may be relatively short as new generations of chips emerge. Meanwhile, competition can place downward pressure upon the prices companies are able to charge for AI services.
The resulting financial structure may prove unstable. Some investments will undoubtedly produce valuable services and profitable businesses. Others may never recover their costs. History offers numerous examples of genuine technological revolutions accompanied by speculative bubbles. Railways, telecommunications and the internet transformed economic life, yet investors in particular enterprises often suffered catastrophic losses. A technology can be revolutionary while the financial expectations surrounding it are irrational.
The political consequences of a major AI investment collapse could be substantial. Governments have increasingly treated AI infrastructure as strategically important, creating pressures for subsidies, regulatory protection and other forms of intervention. If heavily indebted enterprises or financial institutions encounter serious difficulties, political leaders may face demands to prevent wider economic disruption. The resulting interventions could socialise losses while leaving the gains from successful investments concentrated among a relatively small group of companies and investors.
Gore goes further, suggesting that an AI-related financial crisis might become another opportunity for political elites to consolidate power. That possibility should be distinguished from any claim that such a crisis has been deliberately planned. Economic emergencies frequently create opportunities for expanded government intervention without requiring a conspiracy to manufacture them. The relevant political question is whether temporary measures introduced during a crisis become permanent instruments of control.
But the most disturbing implications of AI may arise even if the industry succeeds economically. Indeed, commercial success could make the political dangers considerably greater.
Artificial intelligence is exceptionally well suited to processing enormous quantities of information, identifying patterns, classifying individuals and predicting behaviour. These capabilities have obvious legitimate applications in medicine, logistics, fraud detection and scientific research. They also possess equally obvious applications in surveillance, censorship and political administration.
A government does not require a conscious machine to construct a highly intrusive surveillance system. It requires access to information and the ability to process that information efficiently. AI can assist in analysing communications, identifying social networks, classifying political opinions, detecting behavioural anomalies and assigning risk scores to individuals. Combined with digital identity systems, financial records, facial recognition and centralised administrative databases, such technologies could create mechanisms of social monitoring far beyond the practical capabilities of earlier authoritarian governments.
The essential point is that artificial intelligence does not need to become superintelligent to become politically dangerous. It need only become sufficiently capable of performing the routine tasks of surveillance and classification at an unprecedented scale.
Earlier authoritarian systems depended upon armies of informants, police officers, clerks and bureaucrats. Information had to be collected, filed, interpreted and transmitted through cumbersome administrative organisations. Digital technology has already transformed these processes. Artificial intelligence promises to accelerate the transformation by automating much of the interpretation and classification previously performed by human officials.
The result could be a form of techno-totalitarianism in which political control becomes increasingly automated. Citizens might be classified according to their online activities, financial transactions, associations or perceived political reliability. Access to services could become conditional upon compliance with administrative requirements. Speech restrictions could be enforced through automated moderation systems, while individuals might find it increasingly difficult to challenge decisions produced by opaque algorithms.
Such possibilities do not require assuming that every democratic government intends to establish a dictatorship. Institutions can expand surveillance powers incrementally, often in response to genuine problems. Child protection, fraud prevention, national security and the suppression of criminal activity are all legitimate public concerns. The danger arises when measures introduced for limited purposes become general instruments of monitoring and behavioural regulation.
AI can make this expansion particularly attractive because it reduces the administrative cost of control. Activities that would once have required thousands of officials can increasingly be undertaken by automated systems. The political temptation is obvious. When monitoring becomes inexpensive, governments may become less reluctant to monitor.
There is also a troubling relationship between automated decision-making and political accountability. A human official can be questioned about the reasons for a decision. An automated system may produce a classification or recommendation through processes that are difficult for ordinary citizens to understand and sometimes difficult even for its operators to explain. Responsibility can become dispersed among software developers, government departments, contractors and administrators. The individual affected may discover that no identifiable person is willing or able to provide an adequate justification.
This is one of the most serious threats posed by the uncritical adoption of AI in government. Administrative decisions may acquire an appearance of scientific objectivity while embodying contestable political assumptions. A classification produced by an algorithm can seem more authoritative than an ordinary bureaucratic judgment, even when the algorithm is trained upon incomplete data or designed around questionable criteria.
The same danger exists in the regulation of public discussion. If governments pressure technological platforms to suppress vaguely defined categories of harmful content, automated systems may respond by restricting lawful expression. The operators have incentives to avoid penalties, and the easiest way to minimise regulatory risk may be to remove borderline material. The resulting censorship need not involve a minister personally ordering the suppression of a particular opinion. It can emerge through the interaction of legislation, commercial incentives and automated enforcement.
Gore's warning about the degradation of human intelligence should also be understood in this wider context. A population accustomed to receiving instant answers, accepting automated judgments and allowing machines to perform routine intellectual tasks may gradually lose some of the capacities required for independent citizenship. The danger is not inevitable. Calculators did not abolish mathematics, and computers have greatly expanded the possibilities of scientific inquiry. But technologies can encourage intellectual passivity when convenience replaces understanding.
The decisive question is whether AI remains an instrument under human judgment or becomes a substitute for that judgment. Used critically, it can assist research, expose weaknesses in arguments, accelerate calculations and help individuals investigate unfamiliar subjects. Used uncritically, it can encourage dependence upon systems whose outputs are accepted because they are convenient, fluent and apparently authoritative.
This distinction also exposes a weakness in the more extravagant claims of AI enthusiasts. Even if future systems surpass human performance across many intellectual tasks, it does not follow that their judgments should acquire political or moral authority. Intelligence, however measured, is not equivalent to wisdom, legitimacy or justice. A machine capable of producing exceptionally sophisticated analyses would not thereby acquire the right to determine how human beings should live.
The political problem is therefore deeper than the familiar debate over whether machines will eventually become conscious. A system need not possess consciousness to influence elections, monitor populations, regulate speech or administer access to essential services. Nor does it need genuine moral understanding to enforce rules that governments and corporations have embedded within its operations.
This is why Gore's critique remains important even where his philosophical arguments are overstated. The claim that AI is fundamentally incapable of original thought has not been established. The prediction that the entire industry will collapse is uncertain. But the possibility that AI will strengthen centralised political and economic power is neither remote nor dependent upon speculative breakthroughs in artificial general intelligence.
The greatest irony of the AI revolution may be that the promised liberation from routine intellectual labour coincides with the development of unprecedented mechanisms for supervising human behaviour. The same technology that helps individuals analyse information can help institutions monitor what those individuals say, read, purchase and believe. The same systems that promise to democratise expertise can concentrate informational authority in a handful of corporations and government agencies.
Whether artificial intelligence eventually approaches human intelligence in the fullest sense remains an open question. Whether the present investment boom can justify its enormous financial expectations is another. But the political danger is already sufficiently clear to warrant serious scrutiny.
A civilisation does not need to create a machine more intelligent than humanity to lose control over its own future. It may only need to surrender too much judgment, authority and independence to systems that are efficient at processing information but possess no inherent commitment to human liberty.
The ultimate danger may not be that artificial intelligence becomes too intelligent. It may be that human beings become too willing to obey it.
https://straightlinelogic.com/2025/12/13/ai-is-a-crock-by-robert-gore/