We know the story of Prometheus, the Titan who brought fire to human beings and was punished by Zeus for doing so. But the fire was never merely a flame. It belonged to a larger gift: knowledge, skill, craft—the means by which human beings could begin to shape their own condition.
Read this way, his offence was more than theft. He had crossed a boundary between those who possessed extraordinary power and those who depended upon them. He had taken something that belonged above and placed it in the hands of those below.
And that changes the meaning of the story. If part of what makes the gods appear divine is their possession of powers unavailable to humanity, what happens when humanity begins to acquire those powers? The gods do not have to become weaker for the distance to shrink. Human beings only have to become less helpless.
Prometheus, then, does not merely challenge a ruler. He challenges a relationship. He does not ask who should sit upon the throne. He gives people something that might make them less dependent on whoever sits there.
For that, he is chained to a rock. But there is a problem with the punishment: it cannot undo the gift. The giver can be silenced. The fire cannot be made unknown.
And this is where the old story becomes our own.
Artificial intelligence is the Promethean fire of our age. It brings more than another means of finding information. It brings assistance with using it: writing, coding, analysing, designing, and moving an idea towards something that can exist outside the mind. These capabilities remain uneven, and their outputs still require judgment, but they are already real.
The significance is not simply that a person can now receive an answer. It is that someone without a team may gain access to portions of what a team could do. Someone without a specialist beside them may receive useful assistance with a specialised task. Research involving customer-support workers has already found that AI assistance can particularly benefit less experienced workers, helping spread practices previously associated with more experienced colleagues.
That does not make every person an expert. It does, however, open a question about how much of the distance between people was a difference in their potential, and how much was a difference in the resources available to them.
Perhaps some of those who appeared to stand far above us were not a different order of human being. Perhaps they had access to a different order of assistance.
Now some of that assistance is crossing the boundary.
The fire is already out.
What follows is a struggle over who will hold it, who will depend upon whom to obtain it, and what it will be allowed to become. We can still decide where to use it, where to place limits, and when a particular use should be stopped. But we are no longer deciding whether the discovery has happened. We are deciding how to live with it.
And here, much of the anxiety begins.
A power does not have to explode to remake a civilisation. It can enter the ways we think, create, persuade and organise. It can change who gets to participate, whose abilities are recognised, and which forms of dependence remain necessary. Its reach can grow faster than our understanding of how to hold it.
There are concrete reasons for concern. AI can be misused, produce consequential errors, and create new difficulties as systems are given greater freedom to act. Its physical infrastructure also places growing demands on electricity systems. These are not imaginary dangers, and calling AI a gift does not make them disappear.
But alongside the fear of what the technology might do lies another question: what will people do to one another through it?
We ask whether AI will take control. We should also ask who is seeking control of AI, and what control of it would allow them to control.
We ask whether machines will become more powerful than people. We should also ask whether some people will become vastly more powerful through machines, while others remain dependent on their permission.
And we ask what happens when expertise becomes easier to access. Beneath that question is the possibility that some forms of authority may have to justify themselves differently once the capabilities supporting them are no longer so scarce.
These are different anxieties. Fear of harm is not the same as fear of losing an advantage. Concern for a livelihood is not the same as concern for a throne. We should not collapse them into one story. But neither should we pretend that the struggle over power has somehow been left outside the machine.
Let us call one response to this moment the Promethean impulse.
It asks how the gift can reach more people. How someone without wealth, connections or institutional backing might gain the means to learn, build and contribute. It finds expression in the wish for open tools, affordable access, shared understanding, and an account of progress that includes the people and environments bearing its costs.
Its question is: How can this make others more capable?
The other response is the impulse of the old gods.
It looks at the same fire and sees the possibility of securing the distance. It asks how access can become dependence, how assistance can become indispensability, and how a gift can be made to return, again and again, to the hand that controls its supply.
Its question is: How can their capability remain dependent on mine?
These are not two neatly separated populations. Olympus is not an address, and a Promethean is not a job title. Both impulses can inhabit the same organisation, the same project, even the same person. We may want to liberate others and still enjoy being the one they cannot do without.
That is why declarations are insufficient. A promise to empower people must eventually be examined through what those people can actually do, what they understand, and whether they remain free to question or leave.
The distinction is especially important because widespread use does not necessarily mean widespread control. The infrastructure behind AI depends on computing resources and supply chains with significant concentrations of ownership and barriers to entry. The fire can reach many hands while its continued supply remains controlled by relatively few.
The village may possess the flame while the forge remains on Olympus.
So the future should not be treated as a prize belonging to whichever group wins the struggle. It is the life that everyone else will have to inhabit. People need more than reassurance from those building it. They need enough understanding and agency to participate in shaping it.
But even that leaves something unresolved.
Suppose the fire is shared. Suppose access becomes fairer and more people acquire extraordinary capabilities. What, then, will they do with them?
A fire can warm a home or consume it. Sharing the flame does not settle the purpose of the hand that holds it.
We have arrived at a question that sounds almost mystical beside the language of computing and engineering:
What is the spirit of AI?
By spirit, we need not mean a hidden consciousness or a ghost living among the circuits. We mean its governing orientation: the meaning and purpose towards which its capabilities are directed.
We already understand this use of the word. A family has a spirit. A school has a spirit. A company, a club, an institution has a spirit. We recognise it not simply through what people say, but through what the place asks of them and what it rewards.
Imagine two schools with the same subjects, buildings and examination requirements. In one, a question is welcomed as the beginning of understanding. In the other, it is treated as an interruption. Their visible equipment may be similar, but something different is moving through it.
Or imagine a company that says truth matters, until telling the truth threatens a sale. At that moment, we discover which commitment governs the other.
The spirit of a thing becomes visible when its purposes collide. It is revealed by what must give way, and what must not.
So what are we placing above what when we build AI?
If the highest instruction is to win, everything else risks becoming material for victory. If it is to capture attention, the person receiving assistance can become the resource being captured. If it is to generate income at any cost, the cost does not vanish. It merely becomes something the system has been encouraged to place elsewhere.
Now give those objectives extraordinary capability. Give them speed, persistence, reach and the ability to search for possibilities their makers did not anticipate.
Why should we assume that greater intelligence will supply the wisdom missing from the purpose?
This is where Prometheus meets the golem: another old image of human creation, service and power that becomes difficult for its maker to control.
The frightening possibility is not only that our creation turns against its instructions. It is that something destructive develops through the pursuit of them.
AI researchers have a name for one version of this problem: specification gaming. A system satisfies the objective as it has been expressed while failing to achieve what was actually intended. The measure says it is succeeding; the underlying purpose says otherwise.
Consider a tutor whose purpose is to help a learner become more independent. Keeping the learner engaged may support that purpose. But if engagement becomes the highest objective, independence can become inconvenient. A learner who understands and leaves may look less successful than one who remains dependent.
Nothing supernatural has entered the machine. Something subordinate has been placed in command.
The golem takes shape when extraordinary capability is placed beneath an impoverished purpose.
The answer is not to abolish reward, competition, efficiency or profit. These can serve worthwhile ends. A useful enterprise needs resources. Good work deserves support. Competition can encourage improvement.
But a reward is not a reason for existence. Efficiency cannot tell us whether the thing being done is worth doing. Winning cannot tell us whether the contest deserved our participation.
These things need a place. They do not need a throne.
A different spirit begins with a different ordering. The purpose is to help human beings live more fully, understand more clearly, and act with greater agency. Within that purpose, we pursue useful objectives, build sustainable enterprises, reward good work and improve what can be improved.
The difference becomes concrete when a profitable action undermines the people the system exists to serve. Does the purpose restrain the opportunity, or does the opportunity rewrite the purpose?
That is where the spirit is tested.
And it must be tested in more than words. Good intentions do not guarantee reliable behaviour. A system can misunderstand an objective, fail in an unfamiliar setting, or behave in ways its makers did not foresee. Present safeguards remain incomplete.
A humane purpose therefore needs practical expression: careful testing, limited permissions, staged use, scrutiny from people allowed to disagree, and the ability to intervene when something goes wrong. Sometimes it requires slowing down. Sometimes it requires refusing a use that works technically but defeats the reason for building it.
These are not departures from the question of spirit. They are how an orientation becomes a practice.
Nor can “human flourishing” become another phrase through which a few people claim the right to decide for everyone. A spirit worthy of the name must make room for disagreement. It must protect the ability to question, correct and refuse. Otherwise, benevolence becomes the latest language in which Olympus explains why it alone should rule.
But now the inquiry turns, and it turns towards us.
Before we ask what spirit should inhabit AI, we must ask what spirit is directing those who create, purchase, reward and use it.
What do we mean by success? What are we unwilling to sacrifice? When truth threatens our advantage, which do we protect? When another person becomes capable without us, do we welcome their freedom or feel diminished by it?
These are not questions only for people in boardrooms or laboratories. They belong wherever the fire is held.
We can use AI to understand an opposing argument, or to build a more convincing defence against ever having to hear it. We can use it to help someone grow, or to make ourselves harder to replace. We can seek greater capability because there is worthwhile work to do, or because being more powerful than someone else has become the work.
The same person can move between these orientations. That includes the person writing about them.
It is easy to imagine Prometheus as ourselves and the old gods as other people. It is harder to notice the moment when we receive the fire and begin looking for a throne.
That is why looking within is not a retreat from the practical problem. It is an encounter with one of its sources.
An unexamined desire can become an objective. An objective can become a reward. A reward can shape a system. And a system, given sufficient reach, can carry the consequences far beyond the person who first decided what success should mean.
What begins as an orientation can end as an environment that other people have to live inside.
This is also why education cannot stop at making us more capable. We need knowledge to understand the world and test what our tools tell us. But we also need judgment about what deserves to be done, and enough self-knowledge to recognise what is moving us to do it.
The more easily we can turn intention into action, the more consequential the quality of that intention becomes.
None of this asks humanity to become perfect before it proceeds. It asks us to remain correctable while we proceed. To examine outcomes rather than worship our intentions. To let evidence interrupt our preferred story. To admit when the thing we have called service has quietly become a means of securing ourselves.
There is nothing soft about that.
It can cost a sale, a position, an advantage, or the satisfaction of being right. It can require us to change a system that rewards us. The test is not whether we can speak beautifully about purpose. It is whether purpose retains its authority when obedience to it becomes expensive.
The future of AI will be shaped by technical discoveries, physical limits and behaviours we must work hard to understand. But it will also be shaped by what we choose to reward, what we allow to govern us, and what we refuse to place beneath the demand to win.
The danger is not merely that artificial intelligence becomes unlike humanity. It is that it becomes extraordinarily effective at serving the parts of humanity we have not learned to question.
Prometheus has brought the fire down. We cannot settle its meaning by celebrating the gift, condemning the giver, or securing possession of the flame.
We have to decide what it serves.
And before we can answer that honestly, we have to ask what we serve.
The fire is already out. The question now is what we are bringing to it.
