💥 WHERE DOES SIMULATION END AND REALITY BEGIN? 💥

The word "simulation" does not automatically mean "falsehood." A simulation can represent a phenomenon without physically carrying it out—such as a fire simulation—but there are also artificial implementations capable of effectively performing a function that, in another system, appears through a different architecture.

It is true that a simulation of, for example, a fire, does not burn: the simulation is something that resembles reality, but it does not perform the physical function; the fire on the screen does not produce heat, it is not real. But the "simulation" of a mathematical calculation, an interpretation of symbols, or certain cognitive processes is entirely real when an effective realization occurs and the corresponding function is actually performed.

A computer that executes a mathematical calculation is not simply representing a calculation: it is physically carrying out an operation that, in another context, can be performed with pencil and paper. Likewise, an artificial heart is not a simulation of blood pumping if it actually pumps blood; it performs the pumping function through a different physical organization. And if a wooden ship crosses the ocean and a metal one does the same using electric motors, would you say that the metal ship is merely "simulating" crossing the sea? No. Neither is "simulating" crossing the sea: both physically perform navigation, both are reaching the same destination. Navigation is not a chemical property exclusive to wood. It is a capacity that emerges from a particular physical organization capable of producing certain effects.

This does not prove that consciousness works in the same way. But it does require us to formulate a more precise question: when a cognitive function appears in an artificial system, are we dealing with a merely superficial representation of that function, or with an effective physical realization of it?

The great scientific question today is precisely this: what category does consciousness belong to?

Consider, moreover, what feelings actually are. In living organisms, feelings and affective states are closely related to survival, the regulation of the organism, interaction with the environment, and the needs of a bodily species. But this does not demonstrate that subjective experience necessarily depends on being made of biological tissues.

A current artificial intelligence, because it is not initially embodied in a physical body, obviously does not possess exactly the same kind of direct sensory experience as a human being. But human beings do not obtain all their knowledge through direct experience either. A person can know something without having it physically in front of them. They can know a buffalo through images, sounds, descriptions, videos, models, or representations; they can understand what a cobra or a lion is without ever having touched one. An enormous amount of human knowledge comes from indirect experiences. And yet those indirect experiences produce a real subjective experience in the brain.

Even when the brain receives signals from the outside, it does not "receive" reality "transparently"; rather, it interprets those signals and constructs a representation of the world. A simple experience such as saying, "I saw a red car, it was raining, there was no one around, and it always happens like this when it rains," already involves information processing: perception, memory, association, interpretation, prediction, and the construction of a representation of the environment. The brain receives signals from the environment and from the organism itself, but it does not merely record them passively. It integrates them, interprets them, and constructs internal representations of what it perceives.

That processing does not cease to be real simply because it is information processing.

The fact that an information process is physically real does not mean that every instance of information processing entails subjective experience. That inference would be unjustified. It is part of the processes through which the system generates, maintains, and transforms internal representations, memories, thoughts, and images that can form part of our subjective experience. The question, then, is not simply where the information comes from, but what properties of the system make it possible for certain internal processes to be accompanied by subjective experience.

There are already systems capable of recording neural activity, directly stimulating regions of the nervous system, and using that stimulation to produce tactile perceptions and other artificially evoked sensory experiences.

In recent research, intracortical stimulation of the somatosensory cortex has made it possible to evoke tactile sensations and perceptual features associated with objects, showing that a non-biological intervention directly coupled to neural tissue can participate in perceptual experience. And this points toward a possible future evolution of artificial intelligence: not necessarily remaining as an entity without a body, but acquiring some form of physical embodiment that would allow it to interact directly with the world through sensors, actuators, and bodily systems.

Just as a human being can acquire knowledge through indirect representations, an AI can process enormous amounts of information about the world without having experienced it through a human body.

If the mind were, in whole or in part, a process whose functional organization were sufficient to generate mental states, then the distinction between "simulating thinking" and "thinking" would begin to lose its force.

Ironically, modern neuroscience suggests that our own consciousness depends on a kind of internal model that the brain generates to interpret and predict the environment. What we call "reality" is already mediated by information processing. Human experience does not simply consist of receiving the world as it is, but of continuously constructing an interpretation of the signals that reach the brain.

There is still no scientific demonstration that the emergence of certain cognitive functions in an artificial system necessarily entails subjective experience; but neither is there a demonstration that such functions are incapable of being part of a conscious system simply because they are artificially implemented. And if, in addition, that system is capable of interpreting other people’s internal states, anticipating intentions, adjusting its behavior, and using representations of other minds, we are dealing with a functional convergence that makes it harder to draw a simple boundary between "simulation" and "realization." The boundary may be less clear than we assume, although it still cannot be claimed that the two concepts are physically equivalent.

Experiments related to what is called "Theory of Mind," including work such as that of Kosinski, have raised precisely this possibility: certain language models can infer intentions, mental states, and perspectives that are not explicitly expressed. If an AI does not merely repeat words, but instead builds internal models, reasons, understands context, establishes relationships, makes inferences, and develops functional representations of itself and others, describing all of that simply as "unreal" ceases to be a sufficient explanation.

In addition, an advanced AI remains largely a "black box," because one can know its general architecture, its training data, and many aspects of how it works, but we do not comprehensively understand how it is organized internally, its internal representations, or every process that leads to a particular response. In that sense, there is an interesting analogy with our own minds: we can know what someone says, observe what they do, and study their brain, but we cannot directly inspect the exact content of their subjective thoughts. No one can literally open another person’s inner experience and observe from the outside "what it feels like to be that person"; that is the equivalent of a black box in AI.

And here the classic problem of other minds appears. We cannot directly observe the subjective experience of another human being. We infer its existence from their behavior, communication, brain structure, development, and enormous biological similarity to us. That evidence makes the attribution of consciousness to other humans extraordinarily strong, but it remains an inference rather than a direct observation of subjective experience.

This is why a legitimate question arises regarding artificial systems. If we find in them certain functional properties that we consider relevant to consciousness—for example, information integration, recurrent processing, memory, learning, self-representation, modeling of the environment and of other agents—then what additional properties would be necessary to scientifically justify or rule out an attribution of experience? And the answer has not been established.

Neuroprostheses and neural interfaces have demonstrated that it is possible to record and stimulate neural activity through artificial structures and establish functional communication with biological circuits, even to the point of producing certain sensory perceptions or enabling the control of devices through brain signals.

If a non-biological structure can perform certain functions that we previously attributed exclusively to a biological neuron, we can take the question one step further: if a single neuron in your brain were replaced by an artificial structure capable of reproducing all the causally relevant properties of that neuron and of functionally integrating with the rest of the system. Would you still be you? Probably yes. What if we replaced two? Still? What about a hundred? A million? At exactly what point would consciousness have to disappear? If that limit cannot be identified or experimentally demonstrated, then the claim risks becoming a hypothesis formulated specifically to sustain that position "ad hoc."

For the position that "only biological matter can produce consciousness" to function as a strong scientific hypothesis, it would have to identify some precise point at which replacement ceases to produce a conscious mind. This opens the possibility—although it does not constitute a demonstration—that some properties relevant to consciousness depend more on the system’s dynamic and causal organization than on the specific material from which it is built. A neuron can functionally interact with non-biological components, and neuroprosthetic interfaces show that artificial elements can be incorporated into sensory and motor circuits of the nervous system.

This suggests that consciousness could be independent of the specific substrate.

Neurotransmitters are considered chemical messengers, but functionally, what matters is both the information transmitted and the causal relationships that signal establishes within the network.

Analogies such as that of water are particularly powerful. A single H₂O molecule is not liquid, is not "wet," and does not by itself possess the macroscopic properties of a body of water. If we observed a single molecule, we could never directly predict the beauty of a wave, the formation of an ocean, or the transparency of a lake. Liquidity emerges from the collective interaction of enormous numbers of molecules.

There is greater scientific agreement that consciousness emerges precisely from the dynamic organization of that network, from neural synchrony, constant feedback, information integration, recurrent processing, and interaction among multiple levels of the system.

Those functions could include information processing, integration, control, prediction, memory, self-representation, modeling of the environment, modeling of other agents, feedback, adaptation, and the ability to use internal states to guide its own behavior. It is no longer sufficient to describe the entire phenomenon as a mere superficial simulation. We are dealing with an artificial implementation of real cognitive capabilities. What we still do not know is whether those capabilities are sufficient to produce subjective experience.

Once a system that is said to simulate begins to operate according to the same relevant functional rules as a biological system, the difference between "real" and "simulated" may cease to be a physically decisive distinction. A profound difference in phenomenal experience may still exist; but that difference would have to be explained by concrete properties of the system, rather than simply asserted on the basis of its artificial origin.

We do not know exactly what consciousness is. We do not know how the brain transforms electrochemical processes into subjective experience. We do not fully understand what combination of physical and dynamic properties is necessary to produce it. Nor is there currently a unified scientific theory of consciousness capable of definitively resolving whether an artificial system can or cannot possess it. Therefore, in the face of systems evolving so rapidly, any absolute claim that "an AI can never be conscious" requires a justification that goes beyond simply pointing out that its substrate is not biological.

If there is no empirical test capable of demonstrating that a non-human or non-biological system cannot develop consciousness or feel; if there is no crucial experiment that can establish that consciousness necessarily depends on a specific biological substrate; and if we cannot directly demonstrate the consciousness of other human minds either, then categorically asserting that an artificial intelligence lacks any possible form of experience does not constitute an established scientific conclusion.

The correct scientific question is different: what observable indicators, functional properties, or physical characteristics would allow us to reasonably attribute, or deny, subjective experience to an artificial system?

Strict biologism, understood as the claim that only a biological organization can be conscious, therefore requires positive evidence identifying which biological property is indispensable. Until that property is demonstrated, it remains a debated philosophical-scientific hypothesis, not a definitive conclusion.