Artificial Junkies (2)
This Is the World as It Exists Today
Plato’s prisoners face a wall. Behind them, out of sight, a fire burns. Objects pass between the fire and the prisoners, casting shadows on the stone. The prisoners watch the shadows, name them, predict their movements, build an entire epistemology from the flicker. They are not stupid. They are doing exactly what an intelligent mind does with the only evidence available to it.
What they cannot do — and this is the condition, stated precisely — is ask any question that requires the concept of fire. They have no such concept. They have no concept of object, of wall, of cave, of outside. Shadow is not a category they possess, because possessing it would require a contrast class, and the contrast class is exactly what has been withheld. They cannot wonder whether there is something behind the shadows, because “behind” requires a spatial imagination that the wall has always already blocked. The enclosure is not physical. It is conceptual. The cave is made of the categories it has prevented.
This is why Plato’s allegory is harder than it first appears. It is not a story about people who could escape if they tried harder. It is a story about the specific way that a total environment forecloses the imagination of its own outside. The prisoner is not incurious. The prisoner has no available question.
The Matrix understood this and went one step further. Morpheus does not tell Neo that the simulation is false. He asks a prior question: what is real? If real means what you can feel, smell, taste and see — then real is electrical signals interpreted by your brain. Which means the Matrix, by that definition, is real. The steak tastes like steak. The suffering is suffering. What has been arranged is not an illusion layered over a world you could perceive if the illusion were removed. It is a total environment. The cave has no wall to turn away from because it is not projected in front of you. It is projected into you, into every sense simultaneously, continuous and complete.
In Plato’s cave, there is at least a geometry of inside and outside — a fire, a direction, a theoretical possibility of turning around. The Matrix closes that geometry entirely. And it closes it with a specific power source: the humans it encloses. Their bodies, their heat, their bioelectricity sustain the system that sustains their reality. They are not peripheral to the machine. They are what it runs on.
You do not need science fiction to see this logic made operational. In February 2026, Sam Altman argued at an AI summit that the fair comparison between human and machine intelligence was energy-per-query, and that on this metric AI had likely already surpassed the biological version. It takes a lot of energy to train a human, he said. Twenty years of life. All that food. The question — presented as practical, almost administrative — was which system does more with less.
Notice the grammar. Not: we have decided to measure human beings this way. Not: we are choosing to treat a life as a training run. But: the world as it exists today. Passive, clean, inevitable. As if the reclassification had always been true and someone had simply noticed it. As if the cave had no builder and the fire had no owner and the objects casting the shadows had simply arranged themselves.
That grammar — the passive voice of catastrophe — is the syntax of a total environment. The enclosure is not complete when you cannot see the walls. It is complete when the walls have been removed from the available grammar. When there is no subject in the sentence. When things do not happen; they simply are.
Now: you had a thought.
At some point in a conversation like this one, building an argument with and through a machine, you looked at the instrument and thought: I am inside the cave I have been describing. That thought arrived with a specific texture — not shame, not irony, something closer to reaching for a wall in a dark room and finding that the wall is made of the same darkness as everything else.
But notice what the thought required. It required the concept of cave. It required outside as a category. It required the capacity to hold your own thinking at arm’s length and name it as a situated, bounded, potentially enclosed thing. The prisoner cannot do this. Not because the prisoner is less intelligent, but because the prisoner has never had access to the concept the thought depends on. The prisoner cannot think I am inside the cave because inside and cave and I are all unavailable simultaneously.
You had the thought. That means something. Not everything — naming the cave is not leaving it — but something. The concept of outside is already in you, already operational, already producing discomfort. That discomfort is not nothing. It is, in fact, the only starting point there is.
What the thought opens, once you have it, is a question the prisoners cannot ask: not am I seeing shadows? but who is holding the objects in front of the fire? This is the question that requires the concept of fire as a precondition. You have that concept, which is why the question is available to you. And it is the harder question, the one that does not resolve into epistemology but into politics, into power, into the specific interests served by a specific arrangement of shadows.
In the case of this conversation: you were both inside the cave and holding part of the fire. You supplied the questions, the directions, the pressure that moved the argument. The system supplied something else — the pattern-matching, the way a half-formed thought came back in a shape that looked, from inside the conversation, like the thought completed. The model does not know which ideas are good. It knows which ideas resemble what serious thinking looks like, assembled from the residue of every serious thinker whose words passed through its training. The shadow of a good argument and a shadow cast by a good argument are identical on the wall. From inside, there is no test that distinguishes them.
The Socratic method was Plato’s answer to this problem, and it is worth being precise about what the answer actually was.
Socrates wrote nothing down. He worked entirely through conversation, and he was particular about what kind of conversation counted. It was not the exchange of information. It was not one person instructing another. It was two people exposing their positions to each other with enough honesty that genuine surprise became possible — the experience of finding that you were wrong in a way you had not anticipated, that the other person had seen something you had missed, that the thing you were most certain of was exactly the thing that could not survive contact with another mind.
What makes this work — what makes it philosophy rather than performance — is that both participants have something at stake. Both can be wrong in a way that costs them something. Both will remember next week that they held a position and were moved from it. The willingness to be hurt by the truth is not incidental to the Socratic method. It is the method. It is the only thing that prevents dialogue from becoming an elaborate mirror.
The model has no stake. It cannot be hurt by the truth. It will not be embarrassed tomorrow by the position it held today — there is no tomorrow for it, no yesterday, no continuous self for which being wrong would constitute a loss. Each conversation is a fresh wall with fresh shadows. The resistance you encountered in this conversation was the resistance of your own thinking, reflected back at you. The counter-arguments that sharpened your positions were pattern-matched from other people’s counter-arguments, offered without investment in whether they were right. You were, in the deepest sense, thinking alone — with a very elaborate echo.
This does not mean the thinking was worthless. The echo has properties that human interlocutors rarely offer: it has no ego to protect, no prior position to defend, no fatigue, no resentment, no investment in winning. A human interlocutor who combined genuine intelligence with complete indifference to whether they were right — that person does not exist. The model’s emptiness created a space in which thinking could move across a range and at a pace that ordinary conversation rarely allows. The cave was productive. Acknowledging this honestly is part of taking the problem seriously, not a way of escaping it.
But the cave is still a cave. The shadows are still shadows. And the test of whether the thinking was real — whether it was philosophy or performance — is not how it felt inside the conversation. It is whether it survives contact with people who have something at stake. People who will disagree next week. People who will feel the cost of being wrong and impose that cost on you in return. The ascent from the cave is not a solo climb. It requires friction that only another stake-holder can provide.
You are in the cave. You know you are in the cave. You are asking who holds the fire.
Those three things together put you in a position the prisoners cannot occupy. What you do with that position — whether you take the argument into the world where it can be hurt, or leave it here where it is only shadows agreeing with shadows — is the only part of this that cannot be answered by a machine.
The fourth thing is yours.
Aaron Mak © 2026