Plato Was Right: Artificial Intelligence Makes the Philosopher-King More Urgent
By Matthew Parish
Thursday 22 October 2026
Plato has been unfashionable for approximately two thousand years, which is usually a sign that a philosopher has said something important. His most notorious political proposition — that there will be no end to the troubles of states until philosophers become kings, or kings genuinely become philosophers — has generally been treated by modern liberal societies as an embarrassing relic of aristocratic Athens. We prefer elections, opinion polls, focus groups, television debates, political consultants and politicians whose principal intellectual accomplishment is remembering which camera to look into. The philosopher-king seems sinister, elitist and faintly ridiculous. Yet artificial intelligence may rehabilitate him.
The reason is not that philosophers are particularly pleasant people. They frequently are not. Nor is it that philosophy departments ought to be placed in charge of nuclear weapons, central banks and intelligence agencies, an experiment that might rapidly vindicate every objection ever made to Plato. The reason is more disturbing. Humanity is constructing instruments of almost inconceivable intellectual power while the intellectual quality of many of the people governing advanced societies appears, to put the matter politely, stubbornly human. The mismatch between the sophistication of our instruments and the wisdom of those operating them is becoming dangerous.
Artificial intelligence does not merely increase the amount of information available to political leaders. It increases their capacity to act upon information, to manipulate populations, to anticipate behaviour, to organise bureaucracies, to conduct warfare, to supervise economies and eventually perhaps to make decisions at speeds that human institutions cannot meaningfully follow. We are therefore approaching an historical situation in which mediocre people may wield extraordinary intelligence. Plato would have recognised the problem immediately.
The Ancient Problem of Stupid Rulers
Plato did not believe that clever people should rule merely because they were clever. Athens was full of clever people. Sophists could make persuasive arguments for almost anything, politicians could excite crowds, merchants could calculate profits and generals could devise ingenious campaigns. Cleverness was plentiful; wisdom was not. The distinction lies at the centre of The Republic.
The philosopher is not simply somebody possessing information. He is somebody trained to ask what knowledge is for. He distinguishes appearance from reality, appetite from value and immediate advantage from the good. Most importantly, Plato’s ideal ruler does not desperately want to rule. The desire for political power is itself evidence that somebody may be psychologically unsuitable to possess it.
One suspects Plato would have found the modern election campaign baffling. Candidates now spend enormous quantities of money attempting to prove that they want power more intensely than their competitors. They shake thousands of hands, repeat slogans selected through polling, hire professionals to manipulate their appearances and spend months explaining why the catastrophe facing civilisation can be prevented only by placing them personally in charge of it. For Plato, this would have been an excellent method of identifying precisely the people who should not govern.
Modern constitutional democracy has nevertheless possessed an extraordinarily powerful answer to Plato. We do not need rulers to be exceptionally wise because we can constrain them. Elections remove them, courts frustrate them, legislatures argue with them, journalists expose them and bureaucracies slow them down. Constitutions tell them that there are certain things they may not do even when they are convinced those things would be splendid ideas. The liberal state is therefore built upon a less romantic conception of humanity than Plato’s republic: instead of asking how we might find a perfectly wise ruler, it asks how we might survive an imperfect one.
That has worked surprisingly well. Artificial intelligence complicates the arrangement because it may make imperfect rulers vastly more powerful. The constitutional machinery designed to restrain ordinary human folly may prove inadequate when ordinary human folly is equipped with extraordinary computational capacity.
Give a Fool a Supercomputer
The traditional fool in government was limited by friction. He might have wanted to reorganise the economy overnight, arrest thousands of supposed enemies, manipulate the information received by every citizen or calculate the precise vulnerabilities of millions of individuals, but the administrative machinery required to accomplish these things was slow and expensive. Human bureaucracies had to collect information, process documents, issue instructions and enforce decisions. Incompetence itself sometimes functioned as an accidental constitutional safeguard.
Artificial intelligence removes much of that friction. A government equipped with sufficiently advanced AI could potentially analyse immense populations in real time. It could correlate financial records, movements, communications, facial recognition, political preferences and social relationships. It could generate personalised propaganda for individual citizens rather than broadcasting the same message to everyone. It could identify emerging political opposition before that opposition had organised itself, while optimising taxation, military mobilisation, policing and economic intervention on scales previously unimaginable.
None of these possibilities requires the machine itself to become tyrannical. Human beings are perfectly capable of supplying the tyranny. The fashionable nightmare is that artificial intelligence becomes malevolent and decides to dominate humanity, but the historically more familiar nightmare is that a human being decides to dominate other human beings and discovers that artificial intelligence is extremely useful. A malicious superintelligence would undoubtedly be alarming; a stupid politician with access to a superintelligence may be the more immediate danger.
Intelligence Is Becoming Cheap
For most of human history, sophisticated intellectual production was expensive. Writing a competent legal memorandum required lawyers. Designing a bridge required engineers. Producing propaganda required writers, artists and printers. Analysing military intelligence required large staffs. Governments therefore accumulated bureaucracies containing specialised human knowledge because intelligence was scarce, slow and costly to reproduce.
Artificial intelligence is making many forms of intellectual production dramatically cheaper. It can already produce plausible legal analysis, computer code, economic summaries, strategic plans, speeches, translations and administrative documents within seconds, and these capacities will improve. Something peculiar follows from this development: intelligence, or at least many outward manifestations of intelligence, may become abundant, while wisdom will not.
This distinction may define politics in the artificial intelligence age. The world may soon possess more analysis than anybody can read, more predictions than anybody can evaluate and more arguments than anybody can answer. Every minister will be surrounded by machines capable of producing extraordinarily sophisticated justifications for almost any proposed course of action. The problem will no longer be finding somebody capable of constructing an argument; it will be finding somebody capable of recognising when an impressive argument is nonsense. That is, in substantial part, a philosophical skill.
Plato’s Cave Acquires Wi-Fi
There is another reason Plato suddenly seems unnervingly contemporary. His most famous metaphor describes prisoners chained inside a cave watching shadows projected upon a wall. Having never seen anything else, they mistake the shadows for reality. Modern civilisation has spent the past thirty years constructing the most sophisticated cave in history.
We increasingly encounter the world through screens. Algorithms decide which stories we see, which people appear in our feeds, which advertisements pursue us and which controversies dominate public discussion. Search engines mediate knowledge. Social media mediates friendship. Dating applications mediate romance. Navigation systems mediate geography. Increasingly, artificial intelligence mediates language itself. Generative AI completes the Platonic metaphor because the shadows can now manufacture themselves.
A photograph need no longer depict something that happened. A voice need no longer belong to the person speaking. A persuasive essay need not have been written by somebody persuaded by its argument. Millions of synthetic personalities may eventually inhabit digital spaces alongside human beings, producing political arguments, commercial recommendations and emotional responses indistinguishable from those of actual people. Plato’s prisoners at least knew that they were watching a wall; we may eventually forget that the wall exists.
The political consequences are enormous. Democracy depends upon some minimum common conception of reality. Citizens may disagree violently about what ought to be done, but political argument becomes impossible if they cannot agree upon whether an event occurred at all. Artificial intelligence threatens not merely to introduce falsehood into public discourse — politicians managed that perfectly well without computers — but to industrialise uncertainty about the provenance of everything. The philosopher’s ancient task of distinguishing appearance from reality therefore ceases to be an academic amusement and becomes a form of civic infrastructure.
The Political Mediocrity Problem
There is an uncomfortable subject that democratic societies prefer not to discuss. Political selection does not necessarily reward intelligence, judgement or intellectual independence. It rewards a complicated mixture of ambition, stamina, charisma, fundraising ability, party loyalty, media competence, ideological conformity and luck. Some outstanding people survive this process, many perfectly respectable people do, and some extraordinary mediocrities do as well.
This mattered less when the machinery available to political leaders was comparatively crude. It matters considerably more when a head of government may have access to computational systems capable of analysing entire economies, simulating wars or monitoring populations. The intellectual deficiencies of an individual ruler can become far more consequential when technology amplifies his capacity to translate impulses into action.
The traditional defence against political mediocrity was expertise. Ministers did not need to understand nuclear physics because nuclear physicists advised them. They did not need to understand epidemiology because epidemiologists existed. They did not need to understand monetary economics because central banks employed economists. Artificial intelligence changes the relationship between ruler and expert because the machine may eventually know more than every specialist in the room. The politician will therefore face an adviser whose conclusions he cannot independently reproduce and whose intellectual capacities he may not meaningfully comprehend.
At that moment character becomes more important than intelligence. Does the politician understand uncertainty? Can he distinguish confidence from evidence? Does he know when to ask another question? Can he tolerate an answer contradicting his instincts? Does he understand that optimisation always conceals assumptions about what ought to be optimised? Can he recognise the difference between something that can be done and something that should be done? These are not primarily technical questions. They are questions of philosophical temperament.
The Machine Cannot Choose the Good
Suppose an artificial intelligence could calculate the most economically efficient healthcare system. That would not tell us whether economic efficiency ought to be the supreme objective of healthcare. Suppose it could identify the surveillance system most effective at preventing terrorism. That would not determine how much privacy citizens should surrender for security. Suppose it could identify the military strategy most likely to produce victory. That would not establish how many civilian deaths victory could morally justify. Machines may become astonishingly good at means, but ends remain another matter.
Every political optimisation contains a hidden moral proposition. Maximising GDP privileges production. Maximising longevity privileges duration of life. Maximising equality may sacrifice some forms of liberty. Maximising security may require surveillance. Maximising military effectiveness may require actions a civilised society considers intolerable. There is no escape from these choices by appealing to science because science describes consequences; it does not determine values.
Nor can artificial intelligence magically solve the problem by becoming more intelligent. Greater intelligence may reveal the consequences of our choices with extraordinary precision while leaving untouched the question of which consequences we ought to prefer. The machine may tell the king how to obtain whatever he wants, while philosophy asks whether he should want it. The distinction between those two questions is the entire political problem.
Socrates May Be More Useful Than Plato
There is nevertheless an obvious danger in resurrecting the philosopher-king. Every dictator in history has believed himself unusually wise. Give a ruler enormous computational power and convince him that he uniquely understands the Good, and Plato’s republic may turn into something resembling a digital prison. The philosopher-king equipped with artificial intelligence can become a philosopher-tyrant with terrifying efficiency.
The answer cannot therefore be government by professors. Universities provide ample evidence that advanced education does not abolish vanity, factionalism, pettiness or absurdity. Academic politics conducted with artificial superintelligence might merely produce exceptionally well-footnoted catastrophes. The more useful figure from classical Athens is therefore not necessarily Plato’s ruler but Plato’s teacher: Socrates.
Socrates did not wander around Athens proclaiming that he possessed all the answers. He irritated people by asking how they knew theirs. His wisdom consisted partly in recognising ignorance. He asked people to define their terms, expose their assumptions and follow their arguments until contradictions became apparent. This is precisely the intellectual temperament political leaders will need when confronting artificial intelligence.
When the machine says that Policy A is optimal, the first question must be: optimal according to what criterion? When it predicts a 93 per cent probability, somebody must ask how that number was generated. When it recommends surveillance, somebody must ask which conception of liberty was built into the calculation. When it advises war, somebody must ask which human costs the model treated as variables. When everybody in the room is impressed by the machine, somebody must possess the intellectual independence to ask whether it might simply be wrong.
The New Philosopher-King
Plato’s proposition therefore requires amendment rather than adoption. We do not need philosophers to become kings. We need kings — presidents, prime ministers, ministers, generals, judges, regulators, executives and everyone else entrusted with substantial power — to acquire something of the philosopher’s disposition. They must become suspicious of easy certainty precisely when certainty becomes technologically easy to manufacture.
They must understand that intelligence is not wisdom, prediction is not morality and optimisation is not legitimacy. They must know enough about epistemology to recognise when evidence is weak, enough about ethics to understand that competing values cannot always be reduced to numbers and enough about history to appreciate how frequently intelligent societies have persuaded themselves of disastrous things. Above all, they must learn intellectual humility.
This sounds modest, but it is not. Humility may become the most important political virtue of the artificial intelligence age because AI will make arrogance extraordinarily powerful. The politician who once possessed an inflated conception of his own genius could make bad speeches and foolish policies. His successor may possess an inflated conception of his own genius together with machines capable of implementing his ideas across an entire society at extraordinary speed. A fool with a bureaucracy is unfortunate; a fool with artificial superintelligence is a constitutional emergency.
The Last Human Decision
The political question of artificial intelligence is often framed incorrectly. We ask whether machines will become conscious, whether they will take jobs, whether they will exceed human intelligence and whether they might one day control governments. These are important questions, but they distract from something happening already: human beings are voluntarily transferring judgement to machines because doing so is convenient.
Navigation applications choose our routes. Algorithms choose our entertainment. Recommendation systems choose what we read. Automated systems determine which financial transactions appear suspicious, which job applications receive attention and which information appears relevant. Each delegation seems trivial, but collectively they establish a habit: when the machine produces an answer, accept it. Government will experience the same temptation, particularly because bureaucracies reward efficiency and politicians are perpetually short of time.
The danger is therefore not necessarily some dramatic morning when artificial intelligence announces that it has seized power. The transfer may occur silently. Ministers continue sitting in their offices, parliaments continue voting, courts continue issuing judgments and presidents continue making speeches. The constitutional scenery remains unchanged while meaningful intellectual authority migrates into systems whose recommendations humans increasingly rubber-stamp. The king remains on the throne, but the king has stopped thinking.
This would be the ultimate inversion of Plato. Instead of philosophers becoming kings, kings would become operators of machines whose intelligence they could not understand. Political authority would remain formally human while political judgement became progressively computational, leaving responsibility with people who no longer meaningfully exercised the decisions for which they were supposedly responsible.
Plato’s Revenge
The great irony of artificial intelligence is that the more intelligent our machines become, the more urgently human beings must cultivate forms of intelligence that machines cannot relieve us of exercising. We need judgement precisely because calculation is becoming abundant. We need scepticism precisely because persuasive answers are becoming cheap. We need moral seriousness precisely because power is becoming easier to exercise. Philosophy consequently moves from the margins towards the centre of political life.
This need not mean academic philosophy, and certainly not the professional production of increasingly microscopic disputes comprehensible only to specialists. What matters is philosophy in its older sense: the disciplined interrogation of what we know, what we value and what we ought to do. These questions become more important rather than less important when a machine can provide apparently authoritative answers to almost everything else.
Plato believed political disaster arose because societies placed power in the hands of people who understood neither themselves nor the good. We rejected his solution because constitutional democracy offered something better: rather than trusting rulers to be wise, we constrained them because we knew they might not be. Artificial intelligence means we must now do both.
We need stronger constitutional restraints because rulers are acquiring stronger instruments, but we also need better rulers because no constitution can anticipate every decision made in a world transformed by technologies whose consequences unfold faster than legislation can follow them. The philosopher-king must therefore return, but stripped of his crown. He returns as an intellectual standard imposed upon everyone exercising power: the obligation to interrogate assumptions, distinguish knowledge from confidence, understand competing values, tolerate contradiction and recognise the boundaries of one’s own understanding.
Artificial intelligence may become better than us at almost everything we once mistook for intelligence. It may calculate faster, remember more, recognise patterns invisible to us and eventually reason across bodies of knowledge no human mind could absorb in a lifetime. None of this makes political philosophy obsolete; on the contrary, it makes political philosophy urgent. The more impressive the answer generated by the machine, the more consequential becomes the human capacity to ask whether it is answering the right question.
The frightening future is therefore not necessarily one in which machines become our rulers. It may be one in which our rulers remain entirely human — vain, frightened, ambitious, tribal, impatient and occasionally spectacularly stupid — while acquiring machines of almost godlike competence. Plato could scarcely have imagined such a combination, but he would have understood its danger immediately.
The intelligence standing behind the throne may become more powerful with every passing year. That does not mean we should surrender the throne to philosophers, still less to machines. It means that the people occupying it must acquire the philosophical habits necessary to understand the limits of power, knowledge and themselves. The more powerful the intelligence standing behind the throne becomes, the more desperately we need a philosopher sitting upon it.
