Long before computers, neural networks, or any promise of thinking machines existed, a German philosopher had already anticipated a question that artificial intelligence would reopen two centuries later. Immanuel Kant lived between 1724 and 1804 in the city of Königsberg, which he practically never left. His routine was so precise that, according to accounts from the time, he took the same walk every day at exactly the same time, to the point that local residents used his passing as a reference to know what time it was.
It was from this disciplined life that Kant developed one of the central theses of his philosophy: that knowledge does not result solely from the observation of the world, but also from the way the mind organizes what it observes. Space, time, cause and effect, for him, are not properties that exist ready-made in reality waiting to be discovered, but structures that the mind itself uses to make any experience intelligible. In other words, we do not know the world exactly as it is; we know the world as our capacity for understanding allows it to appear to us.
From this thesis arises the distinction that runs through all of Kant's work: on one side, the phenomenon, which is the world as it appears to us; on the other, the thing-in-itself, which is the world as it is, independent of any observer. Kant was categorical in stating that the thing-in-itself remains, and will remain, out of our reach. But there is an implication of this philosophy that usually goes unnoticed, and which becomes especially relevant in the face of the artificial intelligence era: for knowledge to exist, there must also exist a subject capable of recognizing itself as such, an "I think" that accompanies and organizes all experience. This self-recognition has a particular structure. To think about itself, the mind needs to place itself, even if for an instant, in the position of one observing itself from the outside, as if it were divided between observer and observed. Judging oneself, evaluating oneself, correcting oneself—all of this requires temporarily assuming the point of view of an other, even if this other does not physically exist in front of us.
It is exactly this split that artificial intelligence seems to reproduce, albeit through completely different means. Increasingly autonomous AI systems are designed not only to process external data but to monitor, evaluate, and adjust their own functioning, as if simulating an external gaze upon themselves; this is already enough to ask: if a machine starts to self-evaluate, is it, in some way, replicating this movement between subject and object that Kant identified as the condition for all knowledge? Or is it merely simulating the external signs of this process, without experiencing any of what, for us, is the very act of knowing oneself to exist?
Ultimately, the most disconcerting question of the future may not be whether artificial intelligence will be able to understand the reality in which we live. Perhaps it is another, more uncomfortable one: after centuries of science and philosophy, and now also of increasingly sophisticated algorithms, will we be able to prove that we understand our own reality?0