Where is the world?

AI-assisted translation from the Greek. Greek original

There is a strange experience most people know without giving it much thought: you look at a photograph of yourself and know that it is you. You were there, you remember when it was taken, perhaps you took it yourself; there is no doubt about the identity of the person you see, and yet something in you reacts instantly: that is not what I am like. You do not need to examine the nose, the eyes or the shape of the face to reach this judgement. Usually the judgement has already arrived, and only afterwards do you begin looking for its reason. You look more closely, try to identify what does not fit, but what came first was not analysis. It was a recognition that did not quite come about.

Until recently, this could have remained a minor peculiarity of photography. The lens freezes a fraction of a second, distance changes proportions, the face we know in motion is immobilised in an expression perhaps no one would have had time to see while standing before us. Artificial intelligence, however, has added a possibility that makes the problem harder. It can now generate an image of you from a moment that never happened, in light that never fell on you, in a place you never visited, with a minute alteration of the face corresponding to no photographic exposure. And you may look at this image, which has no existence in history, and without thinking experience the opposite: I recognise myself.

Two conditions thus appear side by side that we would normally expect to agree. On one side is a real photograph, possessing the full truth of its origin but somehow failing to return to a person the presence they recognise as their own. On the other is a synthetic image corresponding to no event, yet capable of producing that recognition. The first can answer “did it happen?” with an unquestionable yes and “is it me?” with a strange no. The second almost reverses the terms. The moment did not happen, but something in the form is recognised.

It would be easy to attribute this to resemblance and leave it there. One image captured the features better, we might say, while the other distorted them. Yet the experience does not seem to operate so analytically. Recognising a face does not usually wait for us to complete a checklist of morphological tests. We do not successively examine eyes, nose, mouth, jaw angle and the distance between eyebrows before finally deciding whom we see. The face reaches us whole. Afterwards we can take it apart, measure its proportions and describe its features, but the first apprehension has already occurred. Something has presented itself as someone before we can explain why.

Artificial intelligence suddenly makes this old operation of perception visible because it can create something we had until now encountered mainly in painting, dreams or imagination: a human face with the particularity of a specific person, without the person bearing it ever having needed to exist. We do not see a generic “man” or “woman”, but someone of a particular age, with a particular gaze, asymmetries, weariness or calm, even that faint impression that something preceded the moment we see and something will follow it. Except there may have been neither a before nor an after. There was only the image.

What exactly exists, then?

The question becomes harder when the synthetic form does not leave us indifferent. A face that may correspond to no human being can evoke fear, attraction, trust or an unexpected sense of repose. It can become associated with music, with a period of one's life, with a difficult night, and years later the sight of it may still carry that relationship. If it is later established that the face was created entirely by an AI model, that information corrects what we can claim about its origin, but it does not travel back in time to cancel what happened when we saw it. Perhaps the person did not exist. But the encounter with their form did.

This does not mean that whatever we see exists because we see it. A dream does not prove that its characters are walking somewhere in the world, just as a synthetic image does not acquire an original with a biography because it moved someone. The same applies to emotion. The fact that something gives me repose does not prove that its object exists outside me in the way I perceive it. But if, afraid of confusing the real with the imaginary, we go to the opposite extreme and say that nothing happened, we have made a different mistake. For the repose occurred, the form was apprehended, the memory took shape, and a real person's lived world changed. The origin may be synthetic without the event of apprehension being synthetic.

Perhaps we need to separate two questions that usually reach us joined together. One is whether something is true. The other is what must already have happened for us to be able to ask whether it is true. The second comes first, not because it matters more, but because truth needs something upon which judgement can be made. To say “this person does not exist”, “this person” must already have existed for me. Denying their external existence does not abolish the phenomenon about which the denial is uttered. On the contrary, it presupposes it.

Here, form precedes the person. Not as some metaphysical entity waiting outside the world to acquire a body, nor as an idea secretly existing before things. It precedes as a possibility of appearance. Before I know who it is, before I know whether they exist, before I establish whether they were photographed or generated, something has already appeared as a person. Truth does not create this appearance after the fact; it comes to examine it.

And then a problem arises that is no longer particularly about artificial intelligence. If a form can appear, be apprehended and enter the lived world without necessarily corresponding to a person somewhere outside us, where exactly does its appearance happen? We could answer that it happens on the screen, but on the screen we find only points of light. Looking closer, we find numerical values, electrical states, circuits and computations. On the other side we find photons reaching the retina, electrochemical activity, neural networks and a brain processing information. All this can be described, and the more we understand it, the more accurate the description becomes. Yet something strange remains: nowhere among these things do we find, as one more object in the chain, the gaze that looked at us.

Pixels do not look. Photons are not peaceful or threatening. A number is not beautiful, and a neuron is not a man standing before us. We need not conclude from this that there is some immaterial region where forms reside. We need only refrain from pretending that we have solved the problem because we have described the material events accompanying it. Somewhere in this process, more than the transmission of information occurred. For the one who saw, a world appeared, and within the world someone appeared.

Perhaps this is where the word “world” begins to become unstable. We use it every day as though we knew exactly what it meant. The world is everything that exists, everything around us, the Earth, the universe, reality. But if one asks not what the world contains but where the world is, the answers soon begin to collapse. It is here. Where is here? On Earth. The Earth is in the solar system, the solar system in the galaxy, the galaxy in the universe. And where is the universe?

The question suddenly seems mistaken, because every place we could offer as an answer is already within what we are trying to locate. If we say the world is everywhere, we have not solved the problem; we have merely turned “where” into “everywhere”. And one can persist with the most childlike and perhaps most difficult question: yes, everywhere—but where?

There need not be an answer. Perhaps the mistake lies in the very demand to locate the world as though it were one more object in a larger space. But then the question's failure reveals something usually hidden: whenever we ask where something is, we have already accepted, without negotiation, a world in which “wheres” exist. The world does not appear as one of its objects. It is already there whenever any object can appear.

And perhaps that is why artificial intelligence, with no intention of doing philosophy, has brought us before such an old question by an entirely new route. Until now, the chief anxiety about synthetic images has been whether we will be able to distinguish true from false. The anxiety is justified and will probably grow. But there is something before it that we risk missing because we rush to verify. Before deciding whether a face is real or synthetic, we have already seen it as a face. Before checking whether a moment happened, the moment has already appeared. Before deciding what stands before us, we are already within a world in which there can be a “before us”.

This does not equate a human being with their image, nor an existing person with a synthetic face. A human being has a body, a history, continuity; they can suffer, desire, refuse, leave, die. We have no reason to attribute any of these things to a face produced by a machine. The difference is enormous and must remain. The strangeness lies elsewhere: this difference becomes available to our judgement after something has already reached us as a form. The decision “this is real” does not come first so that we may then be permitted to see it. First it appears, and then the work of distinguishing begins.

Perhaps this is why a real photograph can fail where a nonexistent moment succeeds. It is not because the second is truer. It is because truth and presence are not synonyms. One concerns, among other things, what happened and what exists independently of our apprehension. The other concerns the fact that something has already reached us and become part of the world we experience. Confusing them is dangerous, but equally dangerous is the demand that the second disappear because it failed the test of the first.

A person who never existed may never acquire a life because someone looked at them. But they may already have acquired a place in a life. And a person who undeniably stood before a lens may look at the exact record of that moment and fail to find in it what they recognise, without thought, as themselves. If these two things can happen, perhaps the age of synthetic forms obliges us not merely to become better at detecting falsehood. It obliges us to look at something that was there long before machines, and that machines—precisely because they can now construct forms with no precedent in the world—make harder to ignore.

Before the question of whether what I see is true, the fact that I see something has already happened. Before the who, there is the this. Before verification, there is appearance. And before we can even ask where what appeared is located, we have already used what we have never managed to locate.

The world.