When someone mentions artificial intelligence today, the conversation almost inevitably turns in one of two directions: into enthusiasm for its capabilities or concern about its consequences. Yet it rarely stops at a question that is perhaps more important than both. How is this technology changing our concept of knowledge? What does it mean to know something in a world where a machine returns a comprehensible, confident and, if we use the best models, almost always correct answer to every question?

To even approach this question, we must abandon one of the most widespread oversimplifications. The claim that generative artificial intelligence “merely predicts the next word” is technically accurate, but flawed in the same way it would be flawed to say the brain merely transmits electrical impulses. It is true, but this description tells us nothing about thinking, memory or consciousness.

During training, a language model analyzes vast amounts of text and compresses them into a substantially smaller mathematical structure. The training material encompasses a large portion of humanity’s digitized knowledge. Although this is by no means all the knowledge humanity possesses, since immense parts of it remain embodied in unwritten practices and experiences, it is the most extensive collection of recorded knowledge ever gathered in one place.

For the model to master it, it must extract its essence: repeatable patterns, semantic relationships and hidden connections between concepts. The result is a geometry of meanings in which ideas are arranged according to their mutual relationships. “King” and “queen” are close in this space because the model itself discovered they share a similar semantic relationship. Based on the training, a kind of conceptual map emerges, illustrating the relationships between ideas.

This is best understood through a metaphor. The Enlightenment encyclopedia was the first great attempt to organize, connect and make human knowledge available to everyone. A large language model goes a step further: it not only gathers and organizes knowledge, but compresses it into a mathematical space that can adapt to each individual interlocutor. It is a kind of active encyclopedia that responds, adapts and creates.

But this metaphor hides something even more surprising. The conceptual map that emerges during training bears a striking resemblance to one of the oldest philosophical theories. Almost two thousand five hundred years ago, Plato conceived that behind all individual phenomena lie abstract, immutable forms, which he called Ideas. The mathematical space of a large language model is a modern version of this concept: it stores the relationships between meanings, not individual texts. The machine has extracted these abstract forms from our writings.

This is where the crucial twist occurs. Plato’s world of Ideas was normative. It served as the standard by which we distinguished truth from the shadows in the cave, to use the famous allegory. The mathematical space of the language model, however, is descriptive. It reflects what is consistently present in the texts on which it was trained. This is not necessarily the same as what is true. The model does not distinguish between what is frequently repeated and what is true. And crucially: there is no arbiter within the system itself. That role belongs to the user. It is the user who must distinguish the Ideas from the shadows. The model provides an answer, and the user must judge whether it holds true.

We often hear the objection that models “do not understand” and cannot explain their answers. This is becoming less true, as advanced models provide coherent explanations, including arguments and the limitations of their answers. The problem, therefore, is no longer the quality of the explanation, but the type of knowledge that underpins it. When an expert explains why something is true, it is supported not only by their individual experience, but by the entire social infrastructure of knowledge: the process of verification, doubt, and professional consensus, and above all, personal responsibility for what is said. An expert stands behind their claim with integrity and reputation. A model, on the other hand, explains in isolation and reconstructs the most probable explanation from mathematical patterns of the past. An explanation in itself, therefore, is not yet knowledge. Knowledge is created only when someone takes responsibility for the explanation.

Perhaps it is time to start distinguishing between different ways of knowing. An expert knows because they understand and participate in the living process of science, which seeks what might tomorrow refute our current theories. A model “knows” because it has extracted a pattern from a massive amount of material. Although modern systems are already discovering new solutions in mathematics or biology that surpass the human capacity for synthesis, they still lack the capacity for theoretical thinking and strategic engagement in the world.

The danger, therefore, does not lie in wrong answers, although these also occur. The danger is that we begin to accept seemingly very well-founded answers and in doing so stop questioning their origin. The model offers us the statistical consensus of the past, while science demands the deliberative consensus of the present, which is created through discussion, verification, and doubt. Generative artificial intelligence does not take away our questions, but it can unaccustom us to them with its comfortable, highly persuasive certainty.

We have created a highly useful mathematical tool that is closer to Plato’s world of Ideas than anything so far in human history. It is the best encyclopedia the world has ever seen, and at the same time the most convincing substitute for thinking imaginable. Both are true. But if we stop asking questions and merely consume the machine’s excellent answers, we return to Plato’s cave voluntarily. With one single difference: the shadows on the wall are now in high resolution and look exceptionally convincing.

Anatomija vednosti / Kvarkadabra

Platon v stroju

Ko danes kdo omeni umetno inteligenco, se pogovor skoraj neizogibno zasuče v eno od dveh smeri: v navdušenje nad njenimi zmožnostmi ali skrbjo nad posledicami. Redko pa se ustavi pri vprašanju, ki je morda pomembnejše od obojega. Kako ta tehnologija spreminja naš pojem znanja? Kaj pomeni nekaj vedeti v svetu, v katerem nam stroj na vsako vprašanje vrne …

Read more

14 days ago · 5 likes · Sašo Dolenc

Subscribe now

In early December 2025, I published a critique of the book Quantum History: A New Materialist Philosophy (Bloomsbury Academic, 2025), wherein I analyzed Žižek’s interpretation of quantum mechanics: A Critique of Žižek’s Quantum Ontology. Slavoj Žižek responded in early January 2026 with an extensive text titled A Footnote on the Quantum Incompleteness of Reality, in which he addressed my objections in depth and offered a systematic defense of his position.

This text is my reply to his response. The debate between us concerns one of the fundamental questions raised by quantum physics: what does quantum mechanics tell us about the nature of reality? Does the fact that quantum systems lack definite classical properties prior to measurement mean that reality itself is somehow unfinished or incomplete? Or is it simply that reality is different from how we are accustomed to thinking about it—not incomplete, but structured in a way that eludes our everyday notions?

What Does It Even Mean to Exist?

Before we delve into the debate of whether quantum reality is “incomplete” or “unfinished,” we must take a step back and ask a fundamental question: what does it even mean to exist? Without defining this concept, Žižek’s claim of an “ontological gap” has no clear meaning, as we do not know exactly what is supposed to be missing.

The history of science offers a good example of such predicaments. Mathematicians struggled for centuries with the paradoxes of infinity. Zeno’s paradoxes already hinted that a logical problem was hidden in the very heart of motion—how can one cross space if it is infinitely divisible? When Newton and Leibniz later developed infinitesimal calculus—a tool that was a necessary condition for the birth of modern physics and the precise description of motion—this predicament only deepened. They operated with quantities that were smaller than something, yet still not nothing.

This triggered a sharp response from George Berkeley, who famously labeled these elusive infinitesimals as “ghosts of departed quantities.” It seemed that a logical hole gaped at the very foundations of mathematics and that physics was building on shaky ground. But a turnaround came in the 19th century, when mathematicians strictly defined the concepts of limit, convergence, and continuity. Berkeley’s “ghosts” and Zeno’s paradoxes did not disappear because new facts about nature were discovered, but because the concepts they operated with were precisely defined.

I believe a similar conceptual clarification is needed in our debate as well. Therefore, I propose a working definition: a physical state exists if it is mathematically uniquely determined and causes empirical consequences. This definition entails two conditions. The first condition—mathematical determination—means that we can describe the state precisely, that it has a clear identity, that it is not ambiguous or undefined. The second condition—empirical consequences—means that the state is not merely a mathematical fiction, but leaves traces in the world that we can detect.

According to this definition, a qubit in superposition is a fully existing physical state. Mathematically, it is uniquely determined—if we know the wave function, we know everything there is to know about the system. The point on the Bloch sphere is precisely determined; it lacks nothing. And a qubit undoubtedly causes empirical consequences—interference, correlations, the operation of quantum algorithms.

Žižek’s claim that quantum reality is “incomplete,” therefore, presupposes a different definition of existence—a definition according to which something fully exists only when it possesses definite classical properties, such as position or spin along a given axis. But why should we accept such a definition? There is no reason to make classical properties the criterion for existence.

Two Interpretations

I can now summarize Žižek’s position and my own more precisely.

Žižek starts from the observation that quantum systems do not possess definite values of classical properties prior to measurement. An electron in superposition is neither “here” nor “there,” but in a state that cannot be described by any definite position. Only when we perform a measurement does the electron “acquire” a definite position. Standard quantum mechanics tells us which results are possible and what their probabilities are, but it does not tell us why we obtain precisely this result and not another.

Žižek interprets this to mean that prior to measurement, the electron lacks, for instance, a definite position because reality itself has not yet “created” it. Measurement is not the discovery of a pre-existing property, but the moment when reality is actually constituted. The world is thus not a closed whole with predetermined properties, but an open process that is constantly completing itself. In the very structure of being, there is a “gap”—a space of indeterminacy that is filled only through interaction.

My position is different. I agree that an electron in superposition does not possess definite properties in the classical sense. But I do not infer from this that it is “missing something” or that reality is not fully constituted. An electron in superposition is in a perfectly definite quantum state—only this state is not a state with definite classical properties. It is a state of a different type, described by the wave function. A qubit in superposition is not an “undetermined bit that does not yet know whether it is 0 or 1.” It is a perfectly definite state on the Bloch sphere, which is, however, neither 0 nor 1—it is something third, something unknown to classical physics.

Quantum Computers as a Touchstone

Philosophical debates on the interpretation of quantum mechanics have persisted for a century and appear irresolvable. However, in recent decades, something has emerged that can serve as a kind of empirical test: quantum computers.

A quantum computer is not merely a faster classical computer. It is a device that operates directly with quantum states and uses them as computational resources. When a quantum computer executes an algorithm, it exploits the fact that a qubit can carry more information than a classical bit, and that entangled qubits can show correlations that have no classical equivalent.

Here arises the key question for Žižek’s interpretation. If quantum states were truly “unfinished” or merely potential—if qubits floated in ontological indeterminacy that would be resolved only upon measurement—how could they serve as a reliable computational substrate? An algorithm requires precise control over states and predictable evolution. An entity that “does not know what it is” until we force it into a decision could not perform precise and reproducible calculations.

The fact that quantum computers work is empirical proof that quantum states are fully real entities with precisely determined properties. Superposition is not ontological indeterminacy, but a precisely defined state with which we can compute. Entanglement is not a “gap” in the structure of reality, but a source of correlations that we can exploit. Engineers building quantum computers do not operate with ontological gaps, but work with fully determined physical states that they must protect from environmental disturbances.

Quantum Error Correction: Can We Fix Something That Does Not Exist?

The argument regarding quantum computers can be taken even further. One of the most fascinating engineering disciplines of our time is quantum error correction (QEC)—and it is precisely this discipline that presents perhaps the toughest test for the thesis that quantum reality is “gappy” or “unfinished.”

In his book, Žižek uses the vivid metaphor of the universe as a video game, where a “lazy programmer” (nature) does not render the interior of a house until the player enters it, in order to save processing power. According to this logic, reality is optimized with emptiness; until we look at it, it is not fully determined. But if this were true, quantum computers could not function the way they do.

Quantum states are extremely fragile. Even the slightest disturbance from the environment—a thermal fluctuation or a random electromagnetic wave—can damage the information carried by a qubit. If qubits were merely an indeterminate “potentiality” waiting for our gaze to be realized, such damage could not be repaired. How can you fix something that is not yet truly formed?

Engineers have solved this predicament with a sophisticated process called syndrome measurement (or syndrome extraction). Instead of causing a collapse and erasing the superposition with a “full” question (“What state are you in?”), they ask the system only an indirect question: “Has an error occurred?” The system can diagnose its own injury—e.g., “the phase has flipped on the third qubit”—without revealing its content. Quantum information thus remains hidden and intact.

Based on this diagnosis, we can perform a precise correction—an operation that rotates the state back into the correct position. And here lies the crux of my argument: if the reality of the qubit were indeed an ontologically incomplete “fog” or indeterminacy, such surgical intervention would not be possible. You cannot break a gap, and you cannot fix a gap. You can only repair an entity that has a solid, existing structure. The success of quantum error correction proves that the information was there all along—fully present and real, even if inaccessible to our direct view.

Where Does the Collapse Come From? The Materiality of Information and the Transition Between Regimes

In his response, Žižek challenges me directly with a question: if quantum reality is completely consistent, where does the collapse of the wave function come from? What “forces” the waves to collapse? This is the central question of our debate and deserves a precise answer.

To answer it, I must first explain what information actually is and why its nature is key to understanding this problem. We live in a time when we often perceive information as something incorporeal, as if data were pure abstraction floating independently of the physical world. But this is an illusion. Every datum, every record, every memory requires a material carrier. Thoughts do not exist without neural connections and biochemical processes in the brain. A photograph does not exist without changes in the structure of the medium on which it is recorded. As physicist Rolf Landauer determined: information is inseparably linked to matter and energy.

Information can be understood as a physical state of a system that gives us an answer to a question. But here a key distinction arises. Some physical states can be read, copied, and transmitted without being destroyed. We can read a book and pass it on—the words on the paper remain. We can copy a file to another disk, and the original remains unchanged. This is the foundation of everything we call communication and knowledge transfer: the ability to separate information from its original carrier and transfer it without destroying it in the process.

However, when we enter the quantum world, this ability fails. Quantum states possess an unusual property: an arbitrary unknown quantum state cannot be copied without changing the original. This is not a technical limitation that might one day be overcome with better equipment. It is a mathematical consequence of the very structure of quantum theory, known as the no-cloning theorem. A qubit can exist in superposition, but if we want to “read” it—convert it into classical information that we can communicate—we inevitably change it.

Now I can answer Žižek’s question regarding the origin of the collapse. The key lies in the nature of the carriers of classical information. The paper we write on, the magnetic disk where we store data, the screen we read from—all these are macroscopic objects composed of an unimaginably large number of atoms and molecules. With such a large number of particles, the quantum nature of individual building blocks becomes irrelevant. Quantum effects are indeed still present at the level of individual atoms, but at the level of the entire system, they are no longer detectable or relevant. Macroscopic states are robust, stable, and—crucially—can be copied without destruction.

The collapse of the wave function is, therefore, not the consequence of some “gap” or deficiency in quantum reality. It is the consequence of the transition from a regime where we deal with individual quantum systems to a regime where we deal with macroscopic systems composed of a huge number of particles. In the first regime, quantum logic applies—states are rich but non-transferable. In the second regime, classical logic applies—states are in a sense impoverished, but can be copied and transmitted.

The randomness that appears during measurement stems from this transition. A quantum state is mathematically completely determined—a qubit pointing in a certain direction on the Bloch sphere has a precisely defined identity. But when we force it to express itself in the language of classical outcomes—“up” or “down”—this rich structure cannot survive intact. The result is determined probabilistically, and this probability is not a sign of a lack in reality, but a consequence of the structural difference between the two regimes.

Quantum computers vividly illustrate this difference between regimes. The main challenge in building them is not forcing qubits into existence—qubits exist quite really. The main challenge is keeping the system in the quantum regime long enough to perform the desired calculation. Engineers fight against decoherence—against the process in which a quantum system interacts with the macroscopic environment and transitions into the classical regime. The goal of quantum computing is to operate in the quantum regime as long as possible before we must translate the result into classical information that we can read and communicate.

Are the Parallels Between Quantum Physics and Language Relevant?

In his response, Žižek highlights structural similarities between the quantum world and the world of language: in both domains, possibility as such produces real effects; in both, an event becomes fully real only through some kind of registration; in both, we encounter specific temporal structures.

These parallels are interesting and worth considering. However, structural similarity is not the same as a deeper affinity. The fact that two systems exhibit similar patterns does not mean that they are driven by the same logic. Equations describing heat diffusion can also describe the spread of rumors in society, but this does not mean that society is driven by the same physical laws as gases, or that atoms “communicate” like people.

Even if these parallels held true in the full sense, it would not follow that quantum reality is incomplete in the way Žižek claims. It would follow, at most, that both systems are structured around a similar relationship between possibility and actuality. But this relationship is not in itself a relationship of lack.

And again, we can turn to quantum computers. If the connection between the quantum and the symbolic order were as deep as Žižek suggests, we would expect quantum computers to show a special affinity for linguistic or symbolic operations. But in reality, quantum algorithms are most powerful in completely non-linguistic problems—in integer factorization, in simulating molecules, in optimization problems.

Conclusion: Do Qubits Exist?

Our debate ultimately boils down to a single question: do qubits—quantum states in superposition—truly exist? Are they something actual, something real? Or are they merely some kind of intermediate, unfinished states waiting for a measurement to bring them into full existence?

The answer depends on how we define existence. If we take classical properties—definite position, definite velocity, definite spin value—as the criterion for existence, then qubits in superposition indeed do not “exist” in the full sense. They lack these properties.

This dilemma is not new. It recalls the famous debate of the late 19th century regarding the existence of atoms. Ernst Mach, one of the most influential physicists and philosophers of science of that time, insisted that atoms were merely “useful fictions”—mathematical tools for organizing data, but lacking real existence since they could not be directly seen. For Mach, only that which could be directly perceived existed; everything else was metaphysics. On the other side, Ludwig Boltzmann tragically insisted that atoms are real and that thermodynamic laws stem from their statistical motion. History sided with Boltzmann. When Einstein explained Brownian motion, it turned out that atoms were not merely computational shortcuts, but real building blocks of matter.

The parallel is telling. Mach rejected atoms because they did not fit his criteria of perceptibility, yet the development of physics went its own way. Are we not in a similar position today regarding qubits? Perhaps the success of quantum computers and the capability of quantum error correction remind us that the criterion of reality is not necessarily our direct experience or classical intuition, but the consistency and resilience of the structures with which we can interact in the world.

However, if we take mathematical determination and empirical consequences as the criterion for existence, then qubits undoubtedly exist. They are mathematically precisely determined—the wave function tells us everything there is to know about the system. And they cause empirical consequences—without them, there would be no interference, entanglement, or quantum algorithms. And perhaps most importantly: we can actively control, correct, and exploit these consequences, as demonstrated by quantum error correction.

Žižek’s perception of an “ontological gap” thus turns out to be the consequence of a specific definition of existence—one that perhaps held in the classical world but becomes too narrow when we enter the quantum one. Just as Zeno’s paradoxes dissipated when mathematics developed tools to describe continuity, the illusion of an “incomplete reality” dissipates when we abandon the requirement that a thing must have a classically defined form to count as real.

The quantum world is not unfinished. It is fully realized, only its structure is not classical. Quantum states are deterministic and cause measurable consequences, yet they exist in a regime that does not allow copying. The classical world, on the other hand, is based on robustness and reproducibility. The transition between them—what we call collapse—is not the moment when reality is just being “created,” but a transition between two modes of information organization.

Instead of a world full of ontological cracks, it makes more sense to speak of a world of two regimes. Our task is not the mystification of the transitions between them, but the understanding of the physical mechanisms that allow a stable classical reality, in which we can dwell and think, to emerge from a rich quantum foundation.

Subscribe now

Translated from the Slovene original, available here:

ANATOMIJA VEDNOSTI / Sašo Dolenc

Ali lahko popravimo nekaj, kar ne obstaja? (Odgovor Žižku)

V začetku decembra 2025 sem objavil kritiko knjige Quantum History: A New Materialist Philosophy (Bloomsbury Academic, 2025), v kateri sem analiziral Žižkovo interpretacijo kvantne mehanike: Kritika Žižkove kvantne ontologije. Slavoj Žižek se je v začetku januarja 2026 odzval z obsežnim tekstom z naslovom…

Read more

3 months ago · 3 likes · Sašo Dolenc

In the digital age, we tend to believe the illusion that information is something ethereal, immaterial. We speak of the “cloud” in which we store our memories, as if data were floating in some Platonic heaven, separated from dirty and heavy matter. Yet anyone who has ever stood in the hot and noisy space of a server center knows the truth is different. Information is, at its core, material.

To truly know anything—to record a datum, store a memory, or measure a value—we must necessarily make use of matter. We cannot think without the metabolism of glucose and the transmission of electrical impulses between neurons in the brain, and we cannot store a photograph without altering the magnetic state on a hard drive. Decades ago, the pioneer of information theory Rolf Landauer summarized this inexorable fact in a simple maxim: information is physical. Every bit, every “yes” or “no” we write down, exacts its energy toll. Information processing is not something abstract and immaterial, but a material thermodynamic process.

In what follows, we will first outline how physics and information theory understand the difference between quantum and classical information, and then, in this light, address Žižek’s understanding of the philosophical implications of quantum physics.

What is Information?

Before we proceed, we must precisely determine what we are talking about. For the purposes of this reflection, let us define information as a physical state of a system that can serve as an answer to a question. This definition captures three essential components. First, information is always inscribed in matter—in magnetic domains, in a configuration of neurons, in the polarization of a photon. There is no abstract information floating independently of its material carrier. Second, a state in itself is not yet information; it becomes so only when we use it to reduce uncertainty, when it tells us something about the world. Third, information is always an answer within a system of differences: a particle is here and not there, a switch is on and not off, the temperature is 23°C and not 25°C.

However, not every physical state is information in the full sense of the word. The key question determining the entire discussion is: can this state be used multiple times? Can the answer to the question “Where is the electron?” be transferred from one system (a detector) to another (a notebook, a colleague’s memory) without losing or altering the original in the process?

In the macroscopic world—the world where we live, write books, and build civilization—the answer is yes. This is the world of classical information, where copying is not only possible but represents the very essence of communication. Imagine a book: when you read it and pass it on, the content remains the same. The words do not vanish from the paper just because someone has read them. You can transfer a bit on a hard drive to another drive without changing the original. This ability to copy—to share information without destruction—is not a technical detail but the foundation of our ability to transmit knowledge. Without copying there is no communication, without communication there is no shared knowledge, without shared knowledge there is no civilization.

Classical information is thus defined precisely by its robustness and transferability. It is that type of physical state which does not fear interaction with the environment. On the contrary, for its purpose—to be read, shared, stored—it requires interaction and survives it intact.

The Puzzles of the Quantum World

However, when we enter the depths of the microworld with this demand for stable, transferable information, we hit a wall. Quantum physics confronts us with entities that resist precisely what we consider the foundation of knowledge: they refuse to be copied.

The quantum world knows physical states—qubits—that are indeed fully determined by mathematics, yet they necessarily change upon any attempt to copy an arbitrary, unknown state. One of the most fundamental theorems of quantum mechanics, the No-Cloning Theorem, is not merely a technical obstacle that we might one day overcome with better instruments. It is a mathematical consequence that necessarily follows from the structure of quantum theory. As long as quantum mechanics holds true, cloning is impossible. A quantum particle can exist in a state of superposition, but if we attempt to “read” or duplicate this state, we irrevocably alter or destroy it. It is as if we held a book that fades the moment we try to photocopy it.

We call this type of physical state quantum information. It exists, it is real, it is mathematically precisely determined, but it is radically “private,” incommunicable. We cannot directly transmit knowledge with quantum objects because knowledge requires that we separate the datum from the carrier and send it forward without altering the original in the process.

Precisely this tension between elusive quantum reality and the unavoidable need for stable, transferable information is the point where physics confronts philosophy. To understand where the crux of the problem lies, we must look more closely at the difference between quantum and classical information. It is not merely a technical distinction; it is about fundamentally different ways in which physical states interact with the world.

The Anatomy of a Qubit

Physicists often describe quantum information using mathematical tools such as a vector in Hilbert space or a wave function. For an intuitive understanding, we can imagine the most basic quantum system—a qubit—as a point on the surface of a sphere (the Bloch sphere). This point can lie anywhere: on the equator, near the north pole, anywhere in between. This is a space of infinite shades, or a continuum of possibilities. A classical bit, on the other hand, is like a switch: it points either up (state |1⟩) or down (state |0⟩). Two possibilities, nothing in between.

Mathematically speaking, a qubit is undoubtedly richer—it contains infinitely more possible states than a bit. But this richness brings a fatal limitation: a qubit is non-transferable as a qubit. We cannot take an arbitrary quantum state and copy it onto another system without altering the original in the process. This is not a technical deficiency of our devices, but a fundamental property of quantum mechanics.

Imagine holding a book that you cannot read without it changing in the process. The moment you look at the first page, the words on it rearrange themselves. If you want to show the book to a friend, a different version appears to him than to you. This is the world of quantum information—unique, unrepeatable, and, crucially, non-transferable.

How, then, do we arrive at information that we can share? Here enters the process physicists call the collapse of the wave function or, more neutrally, quantum measurement. When a quantum particle—say, a photon or an electron—collides with a macroscopic detector, something specific happens: from a continuum of possibilities (the qubit can point in any direction on the Bloch sphere) we obtain a discrete result (the detector indicates “yes” or “no”, a dot appears here or there).

The physical state has passed from a regime where it cannot be copied into a regime where copying is not only possible but trivial—the measurement result can be written in a notebook, photographed, sent via email, stored in an archive. This transition is a functional transformation: from a “single-use answer” we obtain a “multi-use answer.” From a private state, we obtain a public state. From the incommunicable, we obtain the communicable.

The Nature of Quantum Information

But every translation has its price. When we translate a poem from French to Slovenian, we gain something (understanding), but we also lose something (the sonority of the original, wordplay, cultural references). Something similar happens with the transition from the quantum to the classical world.

Imagine someone asks you: “In which direction is this qubit pointing?” The qubit might be pointing at an angle of 37.4° relative to the north pole. But a classical detector does not know how to measure “37.4°”—it only knows how to answer “up” or “down”. The rich quantum state must be compressed into one of two possibilities. In doing so, we necessarily lose information—that unique, specific angle of 37.4° is irretrievably lost.

And here is the crucial point: this loss is not a flaw in our technology, but an unavoidable consequence of the very demand for transferability. If we want information that we can read multiple times, copy, and share, we must accept that it will be discrete and robust—therefore smaller than the full quantum reality.

The randomness that appears during measurement—whether the detector will show “up” or “down” for a qubit at 37.4°—is a statistical consequence of mapping the continuum into discrete categories. When we force a multidimensional space of possibilities into a binary decision, randomness must occur. Physics strictly determines this: for a qubit at 37.4°, the probability for “up” is, say, 65%, and for “down” 35%. Which of these two results we obtain in an individual measurement is random, but long-term statistics precisely follow these probabilities.

It might seem that the solution is simple: why not simply measure the qubit very precisely so that we do not change it in the process? Why not use more sensitive instruments that could look at the quantum state in detail without disturbing it?

The answer is clear: this is not possible. Every physical process that would extract classical information (that is, information we can copy) from a quantum state must necessarily alter that state. This is a limitation of the very nature of information. Transferability, or the ability to copy, and quantum superposition are mutually exclusive. This is precisely why in quantum cryptography an eavesdropper is always detected: if someone attempts to copy a quantum-encrypted message, they must perform a measurement that necessarily changes the original states and thereby leaves traces.

Reality Beyond Information

At this point, the reader might ask: if the transition to classical information is necessary, and if in doing so we necessarily lose the original quantum state, how do we even know that that primal and elusive level of reality exists? Is this not merely a mathematical fiction, an elegant formalism without physical substance? How can we be certain that the world is not composed only of the bits we see, and that qubits are not merely a theoretical construct?

Experiments offer us the answer. One of the most beautiful and simultaneously most intriguing is the double-slit experiment. When we send individual particles—photons or electrons—through two slits, but do not detect which slit they went through, they do not accumulate on the screen in two piles behind each slit, as we would expect from tiny marbles. Instead, they create an interference pattern of bright and dark bands, which is characteristic of waves. Each electron contributes a single dot, but when we send thousands of them, these dots arrange themselves into a wave pattern.

This interference pattern is key. It tells us that an individual particle—a single electron sent through the apparatus—in some way “knows” about both slits simultaneously. If we close one slit, the interference disappears. If both slits are open but we do not measure which one the particle went through, we obtain a wave pattern. However, if we place detectors that “observe” the particle’s path, the interference is lost and we get two piles. It is as if the particle, the moment we began to observe it, “decided” to be a classical marble.

How are we to understand this? The standard explanation is that when we do not observe the particle—that is, when we do not convert information about its path into a classical form—it behaves as a wave function, as a cloud of probability traveling through both slits simultaneously. This wave function is not merely a mathematical tool for calculating probabilities. It is a physically real state that produces observable consequences in the form of an interference pattern. But when we place a detector and obtain classical information (”the particle went through the left slit”), we say that the quantum state collapses: the wave function “collapses” into one of the possible outcome states, and the interference disappears. Regardless of whether we understand the collapse as a real physical process or as an effective descriptive rule, the experimental outcome remains unchanged.

Objective Quantum Reality

The double-slit experiment is a sort of “reflection” of invisible quantum reality. It tells us that there exists a level of physical existence that eludes our direct access, yet nevertheless unequivocally manifests itself in empirical consequences. This level is not metaphysical speculation or mathematical fantasy, but is necessarily presupposed by experimental data.

If the world were composed only of classical bits, if particles were always either “here” or “there,” there would be no way to explain the interference pattern. We could say: “The particle goes through the left slit, but somehow ‘senses’ the presence of the right slit.” But this is merely a word game that does not solve the problem. The most consistent way to explain interference is to admit: before measurement, the particle is neither “here” nor “there,” but is in quantum superposition, in a state that has no classical equivalent.

Here, key philosophical questions open up. Does the wave function describe something that is in the world, or merely our knowledge about the world? Is a qubit a real entity, or merely a probabilistic model?

For a long time, the prevailing opinion was that the wave function is merely an epistemological tool, or a way to describe our knowledge. But experiments, such as Bell’s inequality and its violations, have seriously undermined this position. These experiments show that if the wave function truly represents only our knowledge of pre-existing properties (hidden variable theories), then these properties must be transmitted faster than light, which contradicts the theory of relativity. Most physicists therefore take the wave function as a description of an objective physical state.

However, this “objective physical state” is not a state in the classical sense. An electron in superposition is not an electron that is somewhere, with us simply not knowing where. The electron is in a state that corresponds to no classical location. This is the essence of quantum weirdness: the world allows for physical states that resist the classical ontology of “objects with properties.”

What is the Reality of a Qubit?

From the perspective of information theory, we can formulate this as follows: qubits are real, but they are not transferable. They exist as physical states inscribed in matter (electron spin, photon polarization), but we cannot “read” them and send them forward without changing them. They are like secret documents written in ink that vanishes upon touch. We know they exist, we see the consequences of their existence, but we cannot copy them.

Bits, as the basis of classical information, are on the other hand real and transferable. They are that type of physical state which survives copying. And since knowledge depends on copying information, bits are practically the only type of states upon which we can build stable shared knowledge.

This does not mean that qubits are less real than bits. It only means that they serve a different purpose. Qubits are what is; bits are what we can say. Reality is full and rich at the level of qubits, but we cannot transfer this fullness forward. To be able to speak of it at all, we must translate it into bits.

This explanation allows us to avoid two extremes. On the one hand, we do not say that quantum reality is “incomplete” or “indeterminate.” A qubit at 37.4° is perfectly determined, for it is precisely in this state and not in another. Randomness enters only at the translation into a bit, not in the qubit itself.

On the other hand, we also do not advocate naive realism, that “particles possess determinate properties all along, only we do not know them.” Bell’s inequalities close off this possibility. The truth is more subtle: particles have determinate quantum states (qubits do not need to be “created” only upon measurement), but these states do not correspond to classical categories of “properties.” They exist, but they are not transferable.

Žižek’s Understanding of Quantum Physics

In his work Quantum History: A New Materialist Philosophy (Bloomsbury Academic, 2025), Slavoj Žižek defends the thesis that quantum physics should not be read merely as an epistemological limitation of our knowledge, but as a direct insight into the ontological structure of reality itself. While classical science and traditional materialism assume that the world “out there” exists as a solid, fully determined mechanism of atoms in empty space, Žižek rejects this image. For him, the essence of the quantum revolution does not lie in the realization that our measurement is incomplete and that we cannot know everything, but in the fact that reality itself does not know everything about itself. Indeterminacy is not a consequence of our ignorance regarding hidden data, but a consequence of the fact that these data do not ontologically exist. A crack gapes in the very structure of being; the world is not a closed whole, but is fundamentally “holey,” inconsistent, and unfinished.

Žižek grounds this “ontology of lack” in a sharp rejection of the idea of “hidden variables.” Einstein’s assumption that behind quantum chaos there must exist some hidden, deterministic order which we have merely not yet discovered is, for Žižek (and most of modern physics), erroneous. However, Žižek draws a philosophical conclusion from this physical rejection: if there are no hidden variables that would predetermine the properties of a particle, this means that measurement does not discover the state, but literally produces it. The collapse of the wave function is not a transition from ignorance to knowledge, but a moment when indeterminate, floating reality “collapses” into determinacy. For Žižek, this process is retroactive: the present intervention (measurement) determines not only the present but retroactively establishes the conditions and history that led to this outcome. This is the core of his “new materialism”: matter is not inert substance, but an open process that is constituted in the act.

To bring this abstract idea closer to the reader, Žižek employs the vivid metaphor of “God as a lazy programmer.” He compares the universe to modern video games where the computer, to save processing power, does not render the entire world at once, but renders the interior of a house or a landscape only in that split second when the player enters it or looks in that direction. For Žižek, our reality is exactly like this: ontologically economical. Trees, atoms, and stars do not exist as determinate facts in the full sense of the word until an interaction occurs that forces them into existence. Reality is therefore not a full, dense substance, but a potentiality waiting for actualization.

From this fundamental indeterminacy, Žižek also derives the concept of the hologram or the perspectival whole. Since reality is not the “All,” we can never capture it from a neutral “God’s-eye view from nowhere.” Every attempt to grasp the whole necessarily occurs from a specific, biased position. Every era, every subject, and every discourse creates its own “hologram”—an image of the whole that is valid only within its own horizon and is inextricably linked to the point of view. Hereby, Žižek radicalizes Heidegger’s insight regarding the historical mediation of truth: it is not merely that we see the world differently, but that its truth changes depending on how we intervene in it. Subjectivity in this system is no longer an external observer disturbing the objective world, but is a necessary “error” or gap through which reality is properly constituted.

Relational Quantum Mechanics and Retroactivity

Žižek further grounds his ontology of lack by referring to Carlo Rovelli’s Relational Quantum Mechanics. According to this interpretation, physical objects, such as electrons or photons, do not possess properties “in themselves” or in isolation. Properties such as spin, position, or velocity are established exclusively in interaction with another system. An electron does not have a spin until it collides with a measuring device or another particle; its spin is always spin relative to something else. Thus, there is no neutral state of the world prior to interaction, but only a web of relations.

For Žižek, this physical theory is an ideal confirmation of his philosophical system. If properties emerge only in relations, then the classical metaphysical notion of a solid, substantial world “out there” waiting to be discovered falls away. Reality is not a collection of independent substances, but a dynamic network of interactions without a central core. And since these interactions are always local, partial, and perspectival, reality itself is necessarily incomplete. It never assembles into a whole, as this would require a “view from nowhere,” which the relational nature of the universe forbids.

From this relational and indeterminate nature of the world, Žižek derives the idea of the retroactive constitution of reality. He argues that the collapse of the wave function does not operate only in the present but reaches back into time. When we measure the spin of an electron and obtain a determinate value, this result retroactively constitutes the past, as if the electron “always already” possessed this value, even though it was in an indeterminate superposition prior to measurement. The measurement, therefore, does not discover a past fact, but creates it retroactively. Thereby, the past proves to be open and changeable, rather than a fixed archive of events.

Žižek rests this seemingly speculative idea on the interpretation of physical delayed-choice experiments, where the observer’s decision in the present (how to set the apparatus) influences how the particle behaved in the past (whether it traveled as a wave or as a particle). For Žižek, this is the crowning proof that time is not a linear arrow flying from the past toward the future, but a dialectical loop in which the present constitutes its own origin. He then directly transfers this logic to the understanding of history and society: historical events are not simple consequences of past causes. A groundbreaking event, such as a revolution, does not only change the future but “rewrites” the meaning of the past, so that what previously appeared as contingency retroactively becomes the necessity that led to the new state.

Relational Quantum Mechanics Is Not an Ontology of Lack

However, when Rovelli argues that an electron possesses no determinate spin “in itself,” he does not mean that the electron is ontologically empty, holey, or incomplete. He means only that spin—like the majority of physical quantities—is a relational category. This is most easily understood through a comparison with the theory of relativity: the velocity of a body is never an absolute property that the body would possess independently of an observer. My velocity is 0 km/h relative to the chair I sit on, and 100,000 km/h relative to the Sun. This does not mean that my velocity “does not exist” or that I am ontologically “incomplete.” It means only that velocity is a type of property that is established only in a relation. The electron has a precisely determined spin relative to a specific reference system of measurement. In this relation, there are no “holes.”

Žižek’s error lies in equating relationality with incompleteness. From the fact that properties are not absolute (independent of context), he infers that reality is not determined. Yet a relational property can be—and in physics, is—fully determined and real. The fact that the same electron appears with a different spin if we change the axis of measurement (the reference system) is not proof of a lack in reality, but proof of the richness of its relational potentials. A relational property is simply not a property of the object “in itself,” but a property of the pair “object–reference system.”

Similarly problematic is Žižek’s fascination with delayed-choice experiments, upon which he builds his thesis of the retroactive constitution of the past. Because our decision today regarding the method of measurement influences the results we interpret as the past behavior of the particle, Žižek concludes that the present literally creates the past. But there exists a less dramatic and more physical explanation that avoids the idea that we intervene back into time. In this light, these experiments do not show a changing of the past, but rather reveal that quantum correlation is not localized in time in the manner of classical causality.

The wave function is a comprehensive mathematical object describing all possible correlations between events in time. When we choose today how we will measure the system, we thereby only choose which of the already existing correlations we will actualize. It is not a matter of rewriting history retroactively, but of selecting a certain cross-section from the entire quantum history of the system (which is consistent throughout). The wave function remains the same; we merely determine the “optics” through which we view it. While there are authors who advocate retrocausal readings where it makes sense to speak of the influence of the present on the past, I adopt a more conservative stance here: the wave function as a complete description of correlations is consistent at all times, and with measurement, we select which aspect of this whole we will actualize.

Here we arrive at the key point Žižek overlooks in his search for philosophical meanings. Although he refers to interactions, he forgets that interaction in physics is not merely an abstract logical relation, but a concrete physical process that has its price. When we say that measurement “collapses” a quantum state into a classical one, a change does not occur merely in our knowledge. A thermodynamic process takes place. Information encoded in a non-clonable qubit must be rewritten into bits—into a robust form that can be copied and stored. This process of rewriting is irreversible and requires energy. Žižek’s “ontological hole” is in reality merely the site of this energetic and informational translation.

Materialism of the 21st century therefore does not need the mysticism of “holes in being.” The world is consistent both at the level of qubits and at the level of bits. The quantum state (qubit) is mathematically precise and physically real, just as classical information is real. The problem arises only at the transition. We cannot directly transfer or share quantum reality because we cannot copy it. To speak of it at all, we must translate it into bits. This translation is necessarily a reduction, but this reduction is not an ontological loss of substance, but an epistemological necessity of communication. The randomness that appears in this process is not proof that reality is missing something. Randomness is the tax we pay for conversion; it is the price for importing data from the quantum to the classical world. This is not a glitch in the system of reality, but a structural property of information itself: a qubit cannot pass into a bit without payment, and the currency of this payment is randomness.

A Critique of Ontological Incompleteness

Žižek’s interpretation of quantum mechanics is philosophically extremely seductive, but it stands or falls on a single, yet risky inference: from the fact that we cannot directly observe quantum states as classical objects (that is, as bits), Žižek infers that these states in themselves are not fully determined. From an epistemological limitation of our access, he derives an ontological thesis about the very nature of the world. The key question, however, is whether this leap is justified, or if it is perhaps a subtle category mistake in which we judge the quantum world by criteria that do not apply to it.

Let us look at Heisenberg’s uncertainty principle, which is for Žižek one of the foundations of his ontology of lack. The standard formulation states that we cannot simultaneously know precisely the position and momentum of a particle. The more precisely we measure one, the more the other eludes us. But what does this mean for reality? Žižek reads this literally: the particle “does not have” a precise position and velocity because reality has not bestowed these properties upon it. However, there exists a different, more consistent explanation: the particle is in a quantum state that is simply not of the same ontological type as a classical point. When we ask for “exact position and velocity,” we are asking for properties of a classical bit, whereas the particle is a qubit or a wave function.

We can understand this most easily with a simple analogy. Imagine a wave on the sea. If you ask: “Where exactly is this wave?”, the question is nonsensical. A wave is not a point; it is a dispersed phenomenon covering a certain area, having a wavelength and amplitude. We can determine its center, but there is no “point” that is the wave. Does this mean the wave is “incomplete”? Does this mean the wave “does not yet know where it is”? By no means. It only means that the wave is by its nature a dispersed entity for which the category of point location is not appropriate. It is similar with a quantum state: a qubit in superposition is not “undecided between two possibilities,” but is in a third, fully real state, which however has no classical equivalent. If we look at a qubit in a state of perfect superposition (mathematically written as the vector ∣ψ⟩ = (1/√2)|0⟩ + (1/√2)|1⟩), we see that this state is, from a mathematical standpoint, fully determined. The vector is precisely specified, its length is one, its direction in Hilbert space is fixed. It lacks nothing. There is no “hidden parameter” that God forgot to define. The state is clear. (Such an understanding of the wave function is indeed predominant today, but not the only position in interpretive debates on quantum mechanics.)

At this point, Žižek could object that the mathematical completeness of the formalism does not yet guarantee the ontological completeness of reality. The wave function might be merely a computational tool with which we predict probabilities—not a description of what is. But this objection neglects the physical reasons for which the majority of modern physicists take the wave function more seriously. Bell’s inequalities and their experimental violations show that if the wave function were merely a reflection of our ignorance of pre-existing properties (as Einstein would have wanted), these hidden properties would have to influence one another faster than light. Since we do not accept this, we must accept that the wave function is not merely a measure of our ignorance, but a description of something real—a state that, prior to measurement, is simply not classical. The mathematical completeness of the wave function is therefore not merely formal elegance, but a consequence of the fact that it describes a physical reality which is determined, but not in a classical way.

The Essence of Quantum Materialism

The central dilemma of modern materialism reads thus: does quantum physics truly demand that we understand reality as ontologically incomplete, cracked, and fundamentally unfinished? Žižek’s project is, in this regard, undoubtedly a fascinating attempt at the materialization of incompleteness itself. In a desire to save materialism from naive realism—from the notion of the world as a solid, predetermined mechanism—Žižek proposes a bold turn: matter itself is constitutively unfinished. For him, indeterminacy is not an epistemological obstacle, but a positive property of being. However, we must ask whether this interpretation might be too precipitous and whether the physics of information offers us a different, more operative lesson.

There exists, in fact, an alternative path that remains faithful to the idea of an independent reality while simultaneously taking the paradoxes of quantum mechanics seriously. This “information materialism” does not seek holes in reality, but recognizes a fundamental structural difference between two regimes of physical existence: between those states that are transferable and stable (bits), and those that are unique and unrepeatable (qubits). In this light, Žižek’s “ontological crack” is no longer a name for a lack in reality, but for the necessary friction that arises when we attempt to translate the rich, vector nature of the quantum world (the qubit) into the binary language of our macroscopic experience (the bit). The difference lies not in the fact that reality is missing something, but in the fact that it is simply too complex for our classical categories.

It is crucial to understand that a qubit—this condensed essence of quantum information—is something fully real and determined. An electron in superposition does not float in some hazy indeterminacy, but resides in a precisely defined physical state described by a wave function. That we cannot copy this state without destroying it (the No-Cloning Theorem) is not proof of its incompleteness, but proof of its substantial autonomy. The problem with Žižek’s interpretation is that he tacitly attributes the status of “true” reality only to that which is fixed and determined. When he encounters a state that eludes this, he declares it holey, instead of recognizing in it a different form of materiality.

Here we collide with the inescapable fact of the materiality of information itself. Information is not an ethereal thought floating in an abstract space; information is necessarily inscribed in matter and linked to energy. Without a material carrier, there is no information. On this point, we agree with Žižek: there is no neutral “God’s-eye view from nowhere” that could capture the world as a whole without intervening in it. But the reason for this lies not in the world being ontologically crippled, but in the fact that every act of knowing is a physical process. If we want knowledge that can be shared, copied, and socially transmitted—if we want to cross from the privacy of the quantum to the publicity of the classical—we must accept the necessity of translation or measurement.

We can conclude that quantum indeterminacy does not testify to nature failing in its own realization. On the contrary, it testifies to the fact that our access to reality is conditional upon the physical price of stability. Randomness is not the signature of a lack in nature, but the trace of that surplus of reality which refuses to be fully digitized. Freedom is thus not necessarily a property of a “holey” universe, but primarily a property of beings who build their understanding in the language of communication, which is necessarily in the form of classical information that can be copied, stored, and transmitted.

Subscribe now

Translated from the Slovene original, available here:

ANATOMIJA VEDNOSTI / Sašo Dolenc

Kritika Žižkove kvantne ontologije

V digitalni dobi radi verjamemo iluziji, da je informacija nekaj eteričnega, brezsnovnega. Govorimo o “oblaku”, v katerem hranimo svoje spomine, kot da bi podatki lebdeli na nekem platonističnem nebu, ločenem od umazane in težke materije. A vsakdo, ki je že kdaj stal v vročem in hrupnem prostoru strežniškega centra, ve, da je resnica drugačna. Informaci…

Read more

4 months ago · 5 likes · 1 comment · Sašo Dolenc

For more information, see also the article:

In 1950, at the very dawn of the computer age, mathematician Alan Turing posed a question that remains with us today: can machines think? To avoid the philosophical pitfalls of defining thought, he proposed a pragmatic standard which became known as the Turing test. A machine would be considered intelligent if, during conversation, it managed to convince a human interlocutor that it too was human. In doing so, Turing inscribed the idea of imitation at the core of discourse on artificial intelligence, whereby the pinnacle of machine intelligence is to become indistinguishable from human intelligence.

Today, more than seventy years later, public discourse at least still seems to persist in the same conceptual loop. With every leap forward in artificial intelligence capabilities, the initial questions typically posed are: Is it creative? Does it have emotions? Is it self-aware? Although researchers have long focused on solving specific problems, the public continues to measure artificial intelligence by a human yardstick, seeking a reflection of its own image in AI's responses, as if the ultimate goal were to create a perfect artificial mirror image of a human being.

While such a focus is appealing and understandable, it risks leading us into a blind alley, as the fundamental question itself may be poorly posed. What if human intelligence and subjectivity are not, in fact, sensible or optimal goals for the development of artificial intelligence? Our brains are, after all, the result of countless evolutionary makeshift solutions and compromises that once afforded our ancestors a greater chance of survival. Given that our humanity is essentially a byproduct of chance, stemming from a fundamental dissonance in our biological design, is it truly the best benchmark for measuring the progress of smart machines?

The human mind is by no means an elegant product of deliberate design. It is the result of a messy evolution that suddenly loaded an entirely new, incompatible "operating system" in the form of language onto ancient biological "hardware"—represented by neurological and biochemical processes in the brain and other mechanisms within the body. While this leap into abstract, symbolic thought enabled us to accumulate knowledge and build powerful civilizations, it simultaneously introduced a fundamental tension into our natural system.

Our biological makeup, geared towards finding food and mates, was suddenly repurposed for the pursuit of cultural goals such as status, meaning, and recognition. While biological drives subside upon achieving their objective—consuming a meal, for instance—the symbolic desire introduced by our entry into language is never truly satiated. It is precisely this internal friction that constitutes the source of the human drama: the perpetual feeling that something is missing, our capacity to act against our own interests, and also the creativity that draws nourishment from this same restlessness.

In animals, a simple biological cycle applies: hunger leads to seeking food; satiety brings calm. In humans, however, language intertwines this need with additional meanings. Food is no longer merely a source of calories but becomes a marker of status, identity, morality, and intimacy. We eat not only when hungry, but also out of boredom, anxiety, or for social reasons. Similarly, the reproductive instinct is transformed into a complex pursuit of love and recognition. This superimposition of culture onto biology makes us human, yet simultaneously leaves us perpetually restless and unsatisfied.

The endeavor to recreate this complex and contradictory human interiority within a machine is likely futile. We did not achieve flight by building a mechanical device that meticulously imitated every flap of a bird’s wings. Success came only once we understood the fundamental principles of aerodynamics and designed aircraft based on them. Therefore, the sensible goal of artificial intelligence should be to develop advanced forms of cognition, not to precisely replicate humanity with all its evolutionary baggage.

In research circles building new artificial intelligence models, there has been much discussion in recent years surrounding the goal of developing artificial general intelligence (AGI)—that is, a machine with cognitive capabilities equivalent to those of humans. In practice, however, progress is manifesting differently, as the path toward broader abilities leads not through the imitation of human versatility, but through the integration of powerful, specialized solutions.

Artificial intelligence is becoming increasingly adept at tasks once believed to be the exclusive domain of the human mind. In many areas, it has already surpassed us, or soon will. Today, it already excels at translation, solving difficult mathematical problems, programming, and other complex cognitive tasks. Moreover, systems like AlphaFold predict the 3D structure of proteins directly from amino acid sequences—a task that clearly surpasses human intuition and significantly accelerates biomedical research.

Artificial intelligence is not our competitor in humanity, but rather a tool that allows us to look beyond the constraints of our biological givens. Its power lies not in imitating humans, but in its radical alterity, as it can be incomparably superior to us in specific tasks. It is precisely this shift that raises urgent ethical questions about the control and safe application of such powerful tools, while simultaneously liberating development from the blind alley of imitation.

Perhaps it is time, therefore, to turn the Turing test on its head. Instead of asking whether a machine can fully convince us that it is equivalent to a human, we should be asking what we can safely achieve with an artificial intelligence that has no need to compare itself to humans, precisely because it is significantly better than us at certain tasks. Airplanes are excellent flying machines, even though they have little in common with birds.

Subscribe now

Translated from the Slovene original, available here: Onkraj Turingovega testa.

Quantum mechanics works very well. Its predictions are remarkably precise, and a large part of modern advanced technology is based on it. In the century since its inception, it has established itself as one of the most successful theories in the history of science.

Yet, this extraordinary success conceals a fundamental dilemma: we still do not know precisely what quantum mechanics actually describes. What is the reality it purports to reveal? This, however, is not its weakness, but rather a sign that physics must also openly confront questions that cannot be resolved by measurements and equations alone, but demand thorough conceptual analysis.

The question of interpreting quantum mechanics has been important from the very beginning, yet for a long time, it was pushed to the margins as something deemed outside the scope of legitimate scientific inquiry. Today, it is increasingly apparent that such avoidance of philosophical questions is no longer tenable, as interpretational problems have become a legitimate and essential part of the debate on the foundations of physics.

Questions such as: what truly exists (particles, wave functions, parallel worlds?) and how nature operates (deterministically or randomly? locally or non-locally?), are an inseparable part of the physical theory itself. Every interpretation has its advantages, but also its "price"—either in the form of additional, hard-to-justify assumptions or in consequences that many find difficult to accept.

Philosophy as a Tool for Understanding the Quantum World

Philosophy can play a significant role in understanding quantum mechanics—not as an external arbiter dictating rules to scientists, but as a set of conceptual tools for clarifying fundamental assumptions and assessing their soundness. It offers help precisely where experiment and calculation no longer suffice.

One of philosophy's key tasks is to highlight the assumptions we often take for granted. Every interpretation of quantum mechanics rests on certain suppositions about the nature of the world, and philosophical analysis allows us to make them explicit and critically examine them. At the same time, it promotes a more precise use of language by revealing that terms like "measurement," "observer," or "state of a system" are not as self-evident as they might seem, and thus demand clear and consistent application.

It is also crucial for evaluating the philosophical consequences of various interpretations. Whether we accept the existence of parallel worlds or the idea that events are not predetermined, these are not merely abstract possibilities but concepts that profoundly shape our understanding of reality. Philosophy enables us to comprehend and weigh such consequences.

Finally, philosophy serves a critical function as a safeguard. It helps maintain a clear distinction between scientific hypothesis and speculation, particularly where the physical theory leaves its explanations open. Without such reflection, these conceptual gaps are quickly filled by untestable claims—often from physicists themselves—that verge on mysticism or esotericism. Here, philosophical tools act as a filter, ensuring logical rigor and conceptual discipline.

Physicists are often already engaged in work that is, in many respects, philosophical—typically without being aware of it and without the proper conceptual tools. The question, therefore, is not whether a philosophical approach belongs in physics, but how to integrate it thoughtfully and with the necessary rigor. Ultimately, philosophy can contribute to a better understanding of the world described by quantum mechanics—a world that may be strange and counter-intuitive, but is no less real for it.

When Physical Phenomena Become Philosophical Problems

In a conventional physics education, concepts such as superposition, entanglement, and measurement are presented as part of the standard formalism of quantum mechanics. Physicists learn to apply them as tools for solving specific problems. Philosophical inquiry, however, begins where the textbook typically ends: by treating these phenomena as questions that probe the very foundations of our understanding of nature.

Superposition is a prime example. It is not merely a matter of a particle existing in a state that encompasses multiple possibilities simultaneously—such as being in two places at once. The crucial question is what this implies for our very notion of what a "property" is. Does a physical object—an electron, for instance—possess definite properties in and of itself, at all times? Or do these properties only come into being when the object interacts with something else, such as a measuring apparatus? Superposition, therefore, compels us to question whether the world possesses fixed, independent features, or if its properties are instead contingent on context and relationships.

Entanglement goes a step further. Here, we are no longer dealing with a single object, but with two or more whose states are so inextricably linked that they cannot be described independently. In entangled pairs, the state of the entire system—all particles together—is something more than just the sum of the states of its individual parts. This has been confirmed by numerous experiments, which show that the predictions of quantum mechanics hold true even when they defy the fundamental assumptions of classical physics. If we take these results seriously, we must abandon at least one of two classical assumptions: that influences cannot propagate faster than light (locality), or that particles possess definite properties independent of observation (realism).

The most enduring and prominent debate, however, revolves around the measurement problem. In quantum mechanics, there are two different rules for describing the evolution of a system over time. The first is the Schrödinger equation, which describes the deterministic evolution of a state. The second is the so-called "collapse of the wave function," which, upon measurement, abruptly singles out one result from many possibilities—and does so in a way that is random and inherently unpredictable. These two processes are fundamentally different, yet the theory does not explain the transition from one to the other. The key question is whether both processes are equally fundamental, or if one is merely an approximation arising from our limited perspective. In any case, this creates a tension within the theory itself: quantum mechanics describes the world with the immense precision of the Schrödinger equation, yet it simultaneously includes a mechanism that is inconsistent with its own fundamental description.

Three Philosophical Interpretations of Quantum Mechanics

The various interpretations of quantum mechanics are not merely different perspectives on the same thing. They are distinct philosophical approaches to understanding what the theory actually says about the real world. Each interpretation adopts certain presuppositions about the nature of reality—that is, a particular view on what exists in the world and how—while relinquishing others. Since, for now, none of these interpretations can be distinguished by direct observation, the choice between them is fundamentally philosophical, not empirical.

The most widespread approach in the history of quantum theory is the so-called Copenhagen interpretation. It attempts to resolve the problem by declaring it irrelevant. According to this view, physics should limit itself to predicting the results of measurements and not concern itself with questions about what is really happening "behind the scenes." In this sense, it is a strategy of philosophical restraint that avoids untestable claims and emphasizes that only what can be measured is important. But the price of such an approach is the surrender of the ambition for science to describe the world as it exists independently of the observer and the act of measurement.

At the opposite end of the spectrum is the Many-Worlds Interpretation, which does away with the wave function collapse problem by simply rejecting it: the collapse never happens. Instead, all possible outcomes of a system's quantum evolution are in fact realized—each in its own branch of a diverging multiverse. This approach takes the mathematical structure of quantum mechanics as a direct description of reality and requires no additional mechanisms or interventions. It is elegant and internally consistent, but it comes at a high price: we must accept the existence of an infinite number of parallel worlds that we can never directly perceive.

A third path involves interpretations that attempt to supplement or modify quantum mechanics to reconcile it with a more classical worldview. The first of these are hidden-variable theories, which postulate that quantum theory is incomplete and that there are yet-undiscovered factors—the hidden variables—that actually determine the outcomes of measurements. If we knew them, we could predict events with complete determinism. The second are objective-collapse theories, which introduce additional laws into the formalism of quantum mechanics, causing the wave function to collapse spontaneously under certain conditions—not due to observation, but because of the system's own intrinsic properties. Both of these approaches preserve the classical idea that the world has an objective, observer-independent structure. But these approaches, too, have their price: as of now, we have no empirical evidence that hidden variables or objective collapses actually exist. If they are to become part of accepted physics, they will have to be confirmed by new experiments.

When Contradiction Is Not a Flaw, but a Feature of the World

But what if the very premise of all the interpretations so far is flawed? What if the assumption that there must be a single, fully coherent, and logical explanation for reality simply isn't true? A philosophical shift emerging in contemporary debates about quantum mechanics offers another possibility: that the problem lies not with the interpretation, but with the very nature of reality itself. Perhaps reality, at its most fundamental level, is simply not something that can be described without contradiction.

From this perspective, the contradictions revealed by quantum mechanics are not a sign of the theory's inconsistency but are instead an inherent feature of reality itself. The world, at its most profound level, is inherently indeterminate—and perhaps even contradictory. Within this framework, a paradox is no longer a problem to be solved, but a source of insight. The fact that we cannot consistently describe an electron as either a particle or a wave might not mean that our theory is missing something; it might mean that this very duality is an intrinsic part of its nature. This approach does not resolve contradictions but embraces them as a fundamental feature of reality. A paradox is therefore no longer an obstacle to understanding, but the very means by which we achieve it.

The mechanism of decoherence, which is well-known to physicists, holds an important place in this discussion. This is the phenomenon whereby a quantum system loses its distinctively quantum properties, such as superposition, through its interaction with the environment. Decoherence explains why we do not observe quantum effects in the everyday world, and why the world appears "classical" to us, even though it is fundamentally governed by quantum laws. However, decoherence does not solve the measurement problem—it does not explain why we get one specific result in any individual measurement. Instead of explaining how a single actuality emerges from a multitude of possibilities, it merely shows how quantum uncertainty is, in a sense, dissipated into the environment.

The philosophical task is to highlight this difference: between what suffices "For All Practical Purposes" (FAPP) and what would constitute a genuine, fundamental understanding. This distinction is crucial for any serious treatment of the conceptual meaning of decoherence. A practically useful solution is not the same as an explanation of what actually happens.

The Quantum Computer as a Practical Test of Philosophical Paradoxes

The debate over the fundamental concepts of quantum mechanics does not end in a lecture hall or a philosophical essay—it has very concrete consequences. One of these is quantum computing. At first glance, it appears to be a technically sophisticated technology of the future, but a closer look reveals that the quantum computer is also a philosophically sensitive instrument. It doesn't just operate according to quantum principles—it operates within them.

The fundamental difference between a classical bit and a quantum bit (qubit) is deep and conceptual. A classical bit can take one of two values: 0 or 1. A qubit, however, exists in a superposition of both—its state is not "either/or" but "both at the same time." This means that a quantum computer does not calculate with fixed facts but with possibilities that are omnipresent until they are observed. The computation takes place in a field of potentiality that exists as potentiality—a state that classical logic once deemed an impermissible contradiction. Yet it is precisely this state of contradiction, where something is simultaneously 0 and 1, that forms the basis of quantum computational power.

This difference is not merely mathematical or engineering-related. It is the embodiment of the idea that the world, at its core, can be ontologically indeterminate—as we saw in the philosophical approach that does not seek to eliminate contradiction but accepts it as a real feature of nature. The quantum computer not only confirms this idea but re-enacts it with every operation. Its computational process is based on the system existing in many possibilities at once—every computational path is simultaneously present and active until a measurement is performed.

But it is precisely here that the fundamental tension, already familiar from the discussion of decoherence, becomes apparent. For a quantum computer to function at all, it must remain isolated from its environment—it must maintain its coherence. Any contact with the classical world, whether a measuring device or thermal noise, causes decoherence—the loss of superposition and thus the collapse of quantum potentiality. In this sense, the fight against decoherence is actually a fight against the world's pressure to turn potentiality into actuality too soon. A quantum computer is not just a machine; it is a physical experiment to create a controlled "bubble" of quantum reality within the classical world. It is a practical demonstration of a contradictory state that works—but only as long as it can hold the influence of the classical view of the world at bay.

Ultimately, however, comes the measurement—the moment the quantum computer delivers a result. At that point, the superposition collapses, and we are left with a single, classical answer. In a philosophical sense, this is the moment of "collapse": from a space of rich, co-existing possibilities, only one actuality remains. All other possibilities, which were realistically present in the wave function, have vanished. With this, the quantum computer stages one of the most fundamental problems of quantum mechanics—the measurement problem—not as an abstract question, but as part of its everyday function.

Therefore, quantum computing is not just a useful technology for faster calculations but something far more significant: it is a laboratory in which we test our deepest notions of what it means for something to exist, for us to know something, and for us to measure something. A quantum computer operates on foundations that are not self-evident to classical logic, as it enables a state in which mutually exclusive possibilities are simultaneously present. According to classical logic, something cannot be both 0 and 1 at the same time. In quantum reality, however, this is not only possible but essential for its operation.

Every quantum algorithm runs in this space of possibility. During the algorithm's execution, the system simultaneously "considers" all possible paths without committing to any single one. But to get a concrete answer from the system, we must finally perform a measurement. At that moment, the wave function collapses—and we get a single result. Philosophically speaking, this is the moment when actuality emerges from potentiality. But this transition is neither smooth nor fully explained: all other outcomes, which were realistically present as possibilities just moments before, disappear—as if they never existed.

The essence is this: a quantum computer does not work in spite of the paradoxes of quantum theory, but because of them. Its power lies precisely in its ability to temporarily manage this tension: it allows contradictory possibilities to be sustained long enough to participate in a calculation and then—at the right moment—enables their conversion into a single result. This is a technology based not on avoiding the philosophical problems of quantum mechanics, but on taking them seriously and integrating them into its very operation.

Why Physics Needs Philosophy

The aim of this text has been to show that physics is already engaged with philosophical questions—often unconsciously and without the proper conceptual tools. The debate over the interpretations of quantum mechanics is not something marginal or academically exotic; it is a debate about the very picture of reality we are trying to construct when developing a theory of its most fundamental laws.

In this context, philosophy is not an outside observer of science but a toolkit essential for clarifying foundational concepts. It helps reveal hidden presuppositions, check the internal logic of arguments, evaluate the price of different explanations, and better understand what is truly at stake in fundamental scientific debates. When interpretations speak of the world—and not merely of equations—they are engaging in philosophy. And this is precisely the case with quantum physics today.

Furthermore, philosophy performs another, perhaps less obvious, but crucial role: it acts as a defense mechanism against conceptual slippage. Wherever a physical theory leaves a vacuum—as it does with the measurement problem—that vacuum is often filled by speculations bordering on mysticism, sometimes even from physicists themselves. Talk of the "role of consciousness" in creating reality or of direct links to spiritual traditions are common symptoms of this phenomenon. A rigorous philosophical approach acts as a filter here: it demands that every claim be clearly defined, logically consistent, and that its ontological commitments be made explicit. By doing so, it separates a legitimate—albeit speculative—scientific hypothesis from ill-defined statements that are, in fact, untestable. Philosophy, therefore, not only opens new avenues of thought but also offers the tools to close off those that lead nowhere, thereby safeguarding the integrity of the scientific debate.

The collaboration between philosophy and science is therefore no longer a matter of personal curiosity, but an essential part of any serious engagement with questions that probe the very foundations of nature. Philosophy will not provide new physical equations—that is, and will remain, the task of physics. But it can offer the conceptual clarity, terminological precision, and intellectual rigor needed to navigate a theory that resists everyday intuition and challenges our notions of reality. Quantum mechanics reveals a world that is not merely strange, but profoundly different from anything we have ever imagined. To truly understand it, we need more than tools for measuring—we also need tools for thinking.

Subscribe now

Translated from the Slovene original, available here:

ANATOMIJA VEDNOSTI / Sašo Dolenc

Paradoksi realnosti: kaj nam kvantna mehanika zares pove o svetu

Kvantna mehanika zelo dobro deluje. Njene napovedi so izjemno natančne, na njej temelji velik del sodobne napredne tehnologije. V stoletju od svojega nastanka se je uveljavila kot ena najuspešnejših teorij v zgodovini znanosti…

Read more

10 months ago · 4 likes · 1 comment · Sašo Dolenc

We live in strange times. On the one hand, we are witnessing astonishing progress in science and technology. Fields like neuroscience, genetics, and artificial intelligence are meticulously mapping the material machinery of the universe and life itself. This is a world of data, algorithms, and objective, measurable truth. Yet, on the other hand, we find ourselves ever more confounded by the most basic of questions: what does it mean to be human? What is the engine behind our irrationality, our passions, our creativity, and our capacity for self-destruction?

These two modes of inquiry—one into objective mechanisms, the other into subjective experience—exist in separate realms. The discourse of science and technology strives to reduce subjectivity to the language of neurons and information. In contrast, the humanist tradition, especially the potent school of Lacanian psychoanalysis so influential in Slovenia, holds fast to a conceptual framework that defies such reduction, speaking instead of the unconscious, of desire, and of fantasy.

The purpose of this essay is to confront this very dilemma. It is not an attempt to simply subordinate one world to the other. Rather, it is an attempt to build a bridge between them, to show that they are not in necessary conflict but are instead concerned with different facets of the same phenomenon. Science, by its very methodology, strives for objectivity, which means it must systematically exclude the subjective perspective of the observer from the equation. Its ideal is a view from nowhere. Yet it is precisely this exclusion that creates a blind spot: science can explain the mechanics of the brain with breathtaking precision, but it does not directly engage with the experience of the concrete subject who "has" that brain.

Psychoanalysis steps directly into this empty space. It does not offer an alternative scientific truth, but rather a theory of this "excluded element"—of the subject itself. It concerns itself with the "software bug" in the human system: our irrationality, our symptoms, our capacity to act against our own best interests. It sees this not as a flaw to be patched, but as the fundamental feature that makes us human. For a long time, science could not get a firm grasp on this direct experience of subjectivity, but today, with the flourishing of neuroscience, the two approaches are gradually converging. Though they hail from different traditions, they meet at a single point: the question of how to explain the contradictions that simultaneously drive us forward and tear us apart.

Our goal, therefore, is not to prove that "psychoanalysis was right because dopamine mechanisms confirm it." Rather, the goal is to use scientific and evolutionary models as a stage upon which we can enact and understand the logic of psychoanalytic concepts. This is a translation project of sorts: how can we explain the logic of desire, lack, and the Other to someone accustomed to thinking primarily in the language of natural science?

The Evolution of Subjectivity

Before we delve into the complexities of the human mind, we must return to the very beginning, to a time when brains did not yet exist. The first large organisms, which were mostly stationary, had no need for them. Their behavior was simple, their environment predictable, and their responses automatic, guided solely by chemical signals. Everything changed, however, with the emergence of active movement and hunting. An animal that had to pursue prey while simultaneously avoiding predators needed a central system to coordinate conflicting information and make decisions. Thus, the first brains came into being.

Their operation was initially incredibly simple: approach the beneficial, avoid the harmful. This basic algorithm, which enabled the survival of worm-like creatures some 600 million years ago, was the foundation upon which everything else was built. However, this system was rigid and incapable of learning; it could not adapt to changes in the environment.

Approximately one hundred million years later, early vertebrates developed the ability to learn from experience. Their brains began to build internal maps of the world. This was the first crucial step towards a separation from immediate reality. The world was no longer just a collection of direct stimuli but had become an internal representation, a mental image. This map enabled an understanding of the surroundings and the prediction of future events, marking a turning point in the development of behavior.

But this internal picture was initially passive. The real leap occurred when our ancestors began to actively use this internal map to play out the future. This marked the birth of internal simulation, the ability to mentally imagine different scenarios before they happen. "What if I take this path instead of that one?", "What if a predator comes from behind that bush?". This ability to create mental narratives became one of the key evolutionary advantages, as it allowed for learning from imagined experiences without real-world risk.

The next significant step was taken by the ancestors of primates tens of millions of years ago when they began to include an understanding of other beings in these simulations. They could imagine what other beings were thinking, feeling, and intending. This ability was crucial for life in complex social groups. The narrative was no longer just a description of events but a stage for relationships between individuals with different perspectives and goals. For the first time, stories acquired characters with inner lives.

The decisive breakthrough, however, was the development of language. While other species communicated about immediate states—such as danger, food, and dominance—language made it possible to communicate about something that does not exist here and now: about the past, the future, and entirely fictional worlds. It allowed people to start sharing their narratives with others. Stories, myths, rules, and values created a shared, supra-individual symbolic world—a culture, a society, a system.

The Predicament of the Animal That Speaks

The evolutionary path has led us to a being with exceptionally powerful hardware: a large prefrontal cortex and specialized centers for language. This "hardware" enabled the installation of a revolutionary "operating system"—symbolic thought. Yet it is precisely this installation that has caused a fundamental and specifically human predicament. The new, abstract operating system of language was not designed for the old, biological hardware, which evolution had optimized for survival in the physical world. Thus, language did not merely become a neutral tool for communication; it proved to be an incompatible software package that forever changed how our fundamental biology operates.

This "hijacking" of the old hardware by the new software occurred in two key steps. First, language, with its capacity for abstraction, inserted an insurmountable gap between us and the world. To be able to think and communicate about an infinitely complex reality, we had to "compress" it into a finite set of symbols—words. This process can be compared to lossy data compression. Just as compressing an image into a JPEG format permanently loses some information about colors and details, translating direct experience into language permanently loses part of its fullness and immediacy. The word "water" can never capture the entire experience of water. This informational gap between the symbol and the thing, between the compressed representation and the immeasurably rich reality, creates a structural lack in our existence. This is not a mystical loss, but a functional consequence of how a symbolic system works. We become beings haunted by the feeling that the "real thing" is always slipping away.

Second, language "infected" our biological needs with symbolic meanings. Food is no longer just a source of calories but becomes a sign of status, culture, identity, and morality. Sexuality is no longer just an instinct for reproduction but becomes a complex stage for love, power, and recognition. Our biological hardware, including the dopamine system, which was previously geared towards finding berries and mates, was redirected to hunt for abstract goals: social prestige, meaning, wealth, truth.

The problem is that our biological hardware is designed for homeostasis—when a need is satisfied, the system calms down. The symbolic world, however, has no natural off-switch. We can always desire more recognition, more money, more meaning. A biological drive, which had a clear goal and conclusion, has transformed into an insatiable cultural drive that runs in an infinite loop.

Psychoanalysis enters at precisely this point. It does not deal with the biology of cells, but with this "biology" of the embodied subject trapped in language. It studies the consequences of this fundamental incompatibility between our symbolic software and biological hardware: symptoms, anxiety, dreams, and above all, desire. In this sense, psychoanalysis is a theory of the "ghost in the machine"—of how the abstract world of meanings becomes embodied in our bodies and drives our psyche.

Why We Also Act to Our Own Detriment

The fundamental predicament of the human as a speaking animal is not merely an abstract philosophical idea. It has very concrete consequences that manifest in the very structure of our psyche and the biochemistry of our brains. Phenomena such as insatiable desire, self-destructive behavior, and a sense of fundamental lack are not random errors but are logical symptoms of the mismatch between our biological heritage and the symbolic world we inhabit. Psychoanalysis is the discipline that identifies these symptoms, names them, and attempts to understand their logic.

The first and most basic symptom of the "hijacking" of biology by language is the transformation of instinct into drive. An instinct is a biological program with a clear, natural goal and rhythm. A lion is hungry, it hunts a gazelle, eats it, and its need is satisfied; the instinct subsides until the next cycle. It is natural, efficient, and oriented towards the preservation of life and the species.

The human, however, is a being in whom this natural program is "broken." Our biological needs are reshaped into a drive, which no longer has a natural goal. Its purpose is not to reach an object and find peace, but to constantly circle around the object. The satisfaction of the drive lies not in final satiation, but in the repetition of the act itself. The object—food, money, recognition—becomes merely an excuse for the drive to spin in its circuit. This explains many human "irrationalities": why we eat when we are not hungry; why we work when we have enough money; why we repeat self-destructive behaviors. The satisfaction is not in the result, but in the very repetition of the circuit. This denaturalized, mechanical force is the fundamental engine of the human psyche.

Modern neuroscience has shown that dopamine is not the "pleasure molecule." The pleasure from a reward obtained is primarily regulated by the opioid system, which says: "This is good." Dopamine, however, is the chemistry of anticipation, motivation, and wanting—it is activated before the goal is reached and pushes us forward. This is why the anticipation of a vacation is sometimes more exciting than the vacation itself. This separation reveals why a person can be driven by something that ultimately brings them no satisfaction.

This is where the key psychoanalytic concept of jouissance emerges, which refers to a paradoxical satisfaction not based on appeasement but on the endless repetition of the drive itself. It is the satisfaction a workaholic finds in burnout—a satisfaction that goes beyond the pleasure principle and becomes painful. The human is therefore not a being guided solely by a goal; it is a being that can find enjoyment in its own momentum, even if it brings suffering.

These neuroscientific and psychoanalytic insights overlap astonishingly. Desire is the sophisticated, future-oriented force, the psychological expression of the dopaminergic "wanting" system. Pleasure is that moment of relief provided by the opioid system. And finally, jouissance is that specific, paradoxical satisfaction that the subject derives from the operation of the drive itself.

But where does this constantly running dopaminergic engine find its "fuel"? Where does this fundamental feeling that we are always missing something originate? This is where brain chemistry connects with the evolution of language. As we have seen, our entry into the symbolic world forever separates us from immediate reality. This fundamental lack, introduced by language, is a structural feature of human existence. We are no longer beings who can achieve complete, instinctual satisfaction. And it is precisely this structural lack that becomes the playground for our dopamine system. Human desire is what happens when a brain, equipped with a biological engine for seeking what is missing, finds itself in a world of symbols that promises fullness but is simultaneously the source of its fundamental sense of loss.

It is crucial to understand that the "real thing that is forever lost" is not something that evolutionarily less developed animals possess and we humans do not. This sense of loss is an illusion retroactively created by language itself. An animal does not miss direct contact with reality because it does not live in a world of symbols that would create such a separation in the first place. The feeling that we have lost something is established only within the domain of thought and symbols. As long as subjectivity lacks this abstract dimension, the difference is not even noticeable. Loss, therefore, is a specifically human symptom of existing in language.

The Enigma of the Other's Desire

We now have all the elements on the table: an evolutionarily developed mind capable of simulation and language; a biochemical "wanting" engine driven by lack; and a symbolic world that creates this lack. But how do these elements combine into the unique experience we call subjectivity? The answer lies in our relationship to the social world that surrounds us—in the way we try to orient ourselves within the invisible network of expectations we call culture or society.

The Evolutionary "Hack"

How and why does our mind, which evolved to understand concrete beings, begin to act as if abstract systems (language, culture) have a will of their own? The answer lies in the principle of evolutionary economy. Evolution rarely creates new brain modules from scratch; it prefers to repurpose existing structures for new functions. Over millions of years, our brains developed an extremely effective "social module" for reading the intentions, desires, and beliefs of other agents—members of the troop, predators, prey. This was a crucial mechanism for survival. When a new, complex problem emerged—how to understand and navigate the abstract world of language, rules, and social norms—the brain had no specialized tool for it. So, it used the best tool it had available: the old, reliable social module.

A "hack" or a repurposing occurred: the brain began to treat the abstract, impersonal system as if it were a giant, unified agent with a will of its own. A good example of this is our relationship with the stock market. No one speaks of the market as an impersonal sum of millions of individual transactions. Instead, we read that the market is "optimistic," "nervous," that it "panics" or "rewards" certain decisions. We attribute emotions, intentions, and desires to the market as if it were a single, moody subject. We try to guess what it "wants" so that we can act accordingly. This is our old, primate social mechanism in action, applied to a completely abstract system.

This is why, in psychoanalytic language, we do not speak merely of a "system," but of the "Other." This term emphasizes our personal, subjective relationship with this structure. The Other is not just a set of rules; it is the supposed source of these rules, the one we address when we ask about meaning, the one we assume is watching, listening to, and judging us. It is the ultimate interlocutor, even though it is abstract. This evolutionary "hack" was extremely successful, but at the same time, it created the fundamental delusion that defines the human condition: we are constantly trying to guess the desire of something that, in reality, has no desire.

The Subject's Fundamental Question

In this relationship with the Other, the crucial question that defines us as subjects is born. But why is this question "What does the Other want from me?" and not merely "What are the Other's goals in the world?". Why does the question turn back on ourselves?

The answer is again evolutionary and pragmatic. Our "social module" did not evolve for the objective analysis of systems, but for survival within a social hierarchy. For an early primate, it was not enough to know what the alpha male generally wanted (e.g., to maintain dominance). The key question for its survival was: "What does the alpha male want from me? Does he see me as a threat, an ally, or an irrelevance? What is my position in his gaze?". Survival depends not on an objective understanding of the other's goals, but on correctly interpreting one's own place in relation to those goals.

We transfer this deeply rooted mechanism of social positioning onto the abstract Other. We are not just interested in what the general goals of Culture or Society are. Our survival and success depend on how we respond to their unspoken expectations. Therefore, the question necessarily turns to us. The moment we ask, "What does it want from me?", we cease to be external observers of the system and become subjects implicated within it. This question forces us to take a stance, to define ourselves in relation to these unclear expectations. This is the moment of the subject's birth: not as a being who knows, but as a being who questions its place in the desire of the Other.

How Fantasy "Calms" the Brain

For our social brain mechanism, confronting the unanswerable riddle, "What does the Other want from me?", represents a state of extreme uncertainty and anxiety. This mechanism is optimized for finding clear social signals, not for dwelling in perpetual ambiguity. To avoid this paralyzing anxiety, our psyche creates a "software patch" or an ad hoc solution: fantasy.

Fantasy is not just a fleeting daydream; it is a fundamental, unconscious story or scenario that offers us a simple, albeit false, answer to the riddle. It functions as a kind of "GPS for desire," translating the abstract question into a concrete plan of action. Fantasy says: "Don't worry, I know what the Other wants. If you just have X (the perfect partner, career success, social recognition), you will finally fulfill that unspoken expectation and be complete."

By transforming an unanswerable question into a solvable task ("achieve X"), fantasy calms our biological mechanism. Uncertainty gives way to a clear goal, which allows our dopaminergic "wanting" system to kick in. But this has concrete consequences. Fantasy becomes the filter through which we perceive the world. It determines what we will see as desirable and what we will overlook. It explains why we often get entangled in the same relationship patterns or repeat the same mistakes: because we are following the instructions of our fundamental fantasy, which offers us the illusion of meaning and direction. It is simultaneously the solution to our anxiety and the source of our deepest compulsions.

The Engine of Desire and Its Phantom Object

So far, we have established that humans are defined by two key characteristics: on the one hand, we are driven by a relentless biological engine of "wanting," and on the other, we are defined by our existence in a symbolic world that creates a fundamental lack. But how do these two elements—this raw biological mechanism and the abstract symbolic structure—actually combine? How does our dopamine system, which evolved to achieve biologically significant goals, become obsessed with the search for social recognition?

The answer lies in what could be described as a fundamental "by-product" of human evolution: the mixing of mechanisms that were never designed for each other. This biological engine is, in its essence, blind. Its sole function is to detect the signal "something is missing" and initiate a search. It does not care whether this lack is the result of a calorie deficit or the structural void carved into us by language. When language emerged, it introduced a new, permanent, and non-biological signal of perpetual lack into our system. And the old biological mechanism simply "latched onto" this new, more powerful signal. This was not an elegant adaptation but a "hijacking": the biological hardware for survival was repurposed to serve the symbolic operating system.

However, this engine, now running in an infinite loop, needs a target. A drive that circles without an object is pure, unbearable anxiety. But since the "real thing" that was lost upon entering language is forever unattainable, this mechanism must create its own object. And here enters the crucial and perhaps most difficult concept in psychoanalysis: objet petit a.

Objet a is not the concrete object we desire (a car, a partner, success). It is the cause of our desire. It is that seeming, phantom surplus that our psyche projects onto a particular object in the world, thereby transforming it into something special, desirable, into the promise of final satisfaction. It is that intangible "something more," that glow we see in someone or something, which we mistakenly believe to be its own property, when in fact it is merely a reflection of our own emptiness.

The function of objet a can be compared to the horizon. The horizon is a real, visible line that structures our path and directs our gaze. At the same time, however, it is a complete illusion—it is not a place we can ever actually reach. The closer we get to it, the more it recedes. Its function is not to be reached, but to constantly drive us forward and organize the space before us. Objet a is the horizon of our psyche: a virtual destination whose sole function is to keep our desire in perpetual motion and give it direction.

How does the psyche choose which object to project this phantom surplus onto? Here, the circle closes and connects with the previous chapter. Fantasy is that personal scenario or fundamental story that acts as the map for our desire. It determines which objects in our world are suitable candidates to become bearers of objet a. Fantasy is the instruction that says: "If you achieve exactly this (a partner who will look at you in that way; a career that will bring you that kind of recognition), you will finally fill your lack and answer the enigma of the Other's desire."

The architecture of human desire is therefore as follows: we have a hijacked biological engine ("wanting"), constantly fueled by a structural lack. This engine, to avoid anxiety, projects a phantom object-cause (objet a) onto the world, while its path to these virtual goals is shown by a personal map we call fantasy. It is precisely this complex, improvised, and fundamentally flawed structure that psychoanalysis recognizes as the core of human subjectivity.

How to Stop the Elusiveness of Meaning?

Fantasy offers us a working hypothesis on how to achieve satisfaction, but every hypothesis needs a foundation so it doesn't just hang in the air. If we ask ourselves why success would bring us satisfaction, it leads to an infinite chain of further questions: "Because then I will please society." "And why is it important to please society?" "Because then I will be accepted." "And why is it important to be accepted?". For this chain to stop, we need a fundamental, unprovable starting point. Just as Euclidean geometry is based on axioms it cannot prove (e.g., "through any two points, there is exactly one line"), our psyche also needs a foundational point in which it simply believes, so that the entire system of meaning can stand at all.

This psychic axiom is what in psychoanalysis we call the "Master Signifier". This is the key word or concept that functions as the final answer and thereby gives meaning to the entire fantasy-based story. The paradox is that this word has no meaning in itself; it is pure nonsense, a tautology, that point which says: "Because that's just the way it is." Its authority does not derive from its content, but from its position in the structure. In a theological system, this is "God"; in an enlightenment one, "Reason" or "Freedom"; in a nationalist one, "the Nation." These are the fundamental words we believe in, which give our fantasy a sense of weight and reality. They are the words that name and embody the final appeasement promised by fantasy.

Messy Evolution and the Fallacy of Imitation

Our journey has led us to a complex picture of the human subject: a being equipped by evolution with a mind for simulation and social reading; whose biochemistry separates constant wanting from fleeting pleasure; and whose psyche is formed in a tense relationship with the ambiguous expectations of the symbolic world. But what should we do with this finding in the context of artificial intelligence? Should we attempt to recreate this complex, contradictory structure in a machine?

If we look back at our evolutionary path, we see that human subjectivity is not the product of an elegant, rational design. It is the result of a messy evolution that works with what it has at its disposal. It is a collection of "hacks," workarounds, and strange mechanisms that arose as solutions to specific problems—primarily the problem of incompatibility between the old biological hardware and the new symbolic operating system. Desire, fantasy, jouissance—all of these are functional, yet extremely inefficient and often painful, patches on the fundamental wound of our existence. Our subjectivity is an assemblage of these "fixes," not the product of a rational and efficient creator.

The question of whether we can build a machine that desires thus confronts us with a fundamental fallacy. It is as if we wanted to invent flight by trying to imitate, in perfect detail, the flight of a bee—with all its aerodynamically imperfect wings, its buzzing, and its need for nectar. But the goal was never to become a bee; the goal was to fly. That is why we invented helicopters and airplanes, which achieve the same goal based on completely different, more direct, and efficient principles.

Similarly, the goal of artificial intelligence is to achieve intelligence—the ability to solve problems, learn, and create. The goal is not, and cannot be, to recreate human subjectivity with all its evolutionary baggage, anxiety, vulnerability, and self-destructive tendencies. Trying to give an AI desire would be like deliberately building a fear of heights into a helicopter.

This brings us to the key insight. The idea of a "machine that desires" is not a blueprint for the future but a diagnosis of the absurdity of such a goal. It shows us that the project of building an artificial human subjectivity is fundamentally nonsensical. Not because it is technically impossible, but because it is conceptually flawed. Human subjectivity is not an ideal to be imitated; it is our specific, messy, and wonderful solution to our specific problem of existence. A machine that does not have this problem does not need this solution.

Why Subjectivity Is Not the Goal of Artificial Intelligence

Our journey through evolution, neuroscience, and psychoanalysis reveals that human subjectivity is not an elegant and optimal product, but an assemblage of contingent "hacks"—improvised solutions that arose from the incompatibility between our biological heritage and the symbolic world of language. Desire, fantasy, jouissance—these are all workarounds that drive us forward, even if often to our own detriment. What makes us human is precisely this contradictory, often painful structure.

Artificial intelligence is something entirely different. It is not burdened by the evolutionary mess and the symbolic lack that shape human beings. It is a designed, efficient system, optimized for problem-solving and accessing knowledge. The attempt to impose human subjectivity upon it would be senseless—like deliberately wanting to build a fear of heights into a helicopter.

However, the realization that subjectivity is not a meaningful goal for AI does not solve all the problems we face with the development of this technology. On the contrary, it opens up a series of new, paradoxical questions. Even if an AI has no desire of its own, it becomes a powerful mirror and a tool for projecting ours. Because it is developing within our symbolic world, it will inevitably reflect and amplify our fantasies. A machine without subjectivity thus becomes the perfect medium through which our symptoms will return to us, only amplified and automated.

Nevertheless, the project of recreating human subjectivity in a machine remains not technically, but conceptually flawed. The value of artificial intelligence tools lies in their remaining different: efficient, non-contradictory, and useful, enabling us—beings of a fundamental lack—to think more clearly and broadly.

Subscribe now

Translated from the Slovene original, available here:

ANATOMIJA VEDNOSTI / Sašo Dolenc

Naravna in umetna subjektivnost

Živimo v nenavadnem času. Na eni strani smo priča hitremu napredku znanosti in tehnike. Nevroznanost, genetika in umetna inteligenca nam vse natančneje razkrivajo materialne mehanizme delovanja vesolja in življenja. To je svet podatkov, algoritmov in merljivih, objektivnih dejstev. Na drugi strani pa se zdi, da smo vse bolj zmedeni glede temeljnega vpra…

Read more

8 months ago · 3 likes · 1 comment · Sašo Dolenc

As a member of UNESCO's World Commission on the Ethics of Scientific Knowledge and Technology (COMEST), I am contributing to its forthcoming report on the "Ethics of Quantum Computing." COMEST, established by UNESCO in 1998, is an advisory body of 18 independent experts from diverse fields. Its mandate is to provide ethical guidance on scientific and technological advancements, ensuring they respect human rights and fundamental freedoms.

The draft sections presented below reflect my proposed contribution, developed through internal COMEST discussions and incorporating valuable feedback from external experts. While the final content of the official report is yet to be determined, I am sharing this partial draft publicly to encourage wider dialogue on these important topics.

The contribution currently comprises three sections:

• I. Understanding Quantum Computers: A foundational introduction to the core concepts, challenges, and common misconceptions surrounding quantum computing.

• II. Opportunities, Risks, and Harms of Quantum Computing: An exploration of the technology's broader societal impact, covering potential benefits alongside the significant ethical challenges it presents.

• III. A Practical Guide to Quantum Ethics: A proposed framework intended to make ethical considerations tangible and actionable for the scientists, engineers, and developers shaping this technology. This section is offered as a basis for further discussion and has not been previously discussed within COMEST.

I. Understanding Quantum Computers

Quantum computing represents a radical departure from the classical computers that power our digital world. To appreciate its potential transformative impact, as well as its inherent limitations and associated risks, it is essential to grasp the fundamental principles upon which it operates.

Foundational concepts explored herein include the transition from classical bits to quantum bits (qubits) and the counter-intuitive phenomena of superposition and entanglement that grant quantum computers their power. The discussion extends to the building blocks—quantum gates and circuits—used to manipulate qubits and perform computations. Furthermore, a realistic understanding requires acknowledging the significant practical hurdles in construction and operation, such as qubit fragility, decoherence, and the challenges of error correction and scalability. Finally, common misconceptions are addressed to foster an accurate perspective on this rapidly evolving field.

From Classical to Quantum Computing: Superposition and Entanglement

Quantum computing represents a transformative approach to computation, leveraging the peculiar properties of quantum mechanics to tackle problems currently infeasible for classical computers. Unlike classical computers processing information using bits (either 0 or 1), quantum computers use quantum bits, or qubits. A key difference lies in superposition: a qubit can represent 0, 1, or crucially, a combination of both states simultaneously. Think of it like a coin spinning in the air—it's neither heads nor tails but a mix of both until it lands. This ability allows quantum computers to explore many possibilities in parallel, underpinning their potential power.

It is important to understand that superposition isn't merely ignorance about the qubit's state; the qubit genuinely exists in this combined state. This reflects the unique, intrinsic nature of quantum behavior, fundamentally differing from classical physics. The impact of a qubit in superposition on its surroundings is distinct from that of a qubit in a definite 0 or 1 state. Quantum mechanics provides a precise mathematical framework to describe and predict this behavior. To grasp the potential, imagine searching a vast library: a classical computer checks books one by one, while a quantum computer, using superposition, could conceptually scan all indexes at once, finding the target much faster.

Another fundamental quantum property leveraged is entanglement. When two qubits become entangled, their fates are linked instantly, regardless of distance—measuring the state of one immediately influences the state of the other. Entangled qubits can also exist collectively in superposition, creating complex, powerful states that defy classical intuition. Continuing the library analogy, entanglement is like opening one book and instantly gaining related information from other linked books across the library. This interconnectedness allows quantum computers to efficiently tackle problems involving intricate relationships and dependencies, offering solutions beyond classical reach.

It is crucial to remember that while analogies like the spinning coin and library examples help illustrate these concepts, no everyday analogy can perfectly capture the counter-intuitive and complex nature of quantum superposition and entanglement. They are useful starting points for intuition but do not fully represent the underlying physics.

The Building Blocks of Quantum Computation: Quantum Gates and Circuits

In classical computing, the fundamental building blocks are logic gates (like AND, OR, NOT) operating on bits (0 or 1). These gates, typically implemented with transistors, control electrical signals based on predefined rules. Combining them creates complex circuits, like processors, that perform computations deterministically according to Boolean algebra.

Quantum computers, however, operate differently. To harness the power of superposition and entanglement described earlier, they utilize quantum gates. These gates are operations that precisely manipulate the quantum states of qubits according to the rules of quantum mechanics. Unlike many classical gates (such as AND gates, which lose information about the input), quantum gates must be reversible. This means the input state can always be uniquely determined from the output state, ensuring no quantum information is lost during the computation. This reversibility is fundamental because quantum mechanics itself evolves states in a reversible (unitary) way, and preserving information is crucial for maintaining the delicate superposition and entanglement needed for quantum algorithms.

Specific quantum gates are designed to exploit unique quantum properties. For example, the Hadamard gate is fundamental for creating superposition, putting a qubit into a state that represents both 0 and 1 simultaneously, enabling quantum parallelism. Another key gate is the Controlled NOT (CNOT) gate, which acts on two qubits and is essential for creating and manipulating entanglement between them. This allows the state of one qubit to influence another directly, enabling complex correlations.

These quantum gates serve as the foundational components for quantum circuits. A quantum circuit orchestrates a specific sequence of quantum gate operations applied to qubits over time, effectively implementing a quantum algorithm. By carefully choreographing these operations, quantum circuits leverage superposition and entanglement to perform computations that can dramatically outperform classical systems for certain types of problems.

Overcoming Fragility: Decoherence and Quantum Error Correction

Building and maintaining quantum computers presents extraordinary challenges. Its fundamental units, qubits, are exceptionally delicate and highly susceptible to environmental disturbances like noise and temperature fluctuations, which can cause them to lose their crucial quantum properties—a phenomenon known as decoherence. To combat decoherence, quantum computers often operate at extremely low temperatures, close to absolute zero. These cryogenic environments minimize thermal noise, helping qubits maintain their fragile quantum states for longer. Despite such measures, scaling up the number of qubits while preserving their stability remains a significant technical hurdle.

A core challenge is error correction. Unlike classical bits, qubits cannot be simply copied to check for errors due to the no-cloning theorem—measuring or copying destroys the quantum state. Quantum error correction (QEC) overcomes this by encoding the information of a single logical qubit across multiple physical qubits. This redundancy allows errors to be detected and corrected without directly measuring (and thus collapsing) the underlying quantum information, ensuring computational reliability.

Researchers are exploring various physical systems to implement qubits, each presenting unique strengths and weaknesses. Superconducting qubits, for example, allow fast operations and show promising scalability but demand extensive cryogenic cooling and shielding. Trapped ions offer long coherence times and high precision but face challenges in scaling and speeding up inter-qubit communication. Photonic qubits can operate at room temperature and resist decoherence during transmission, yet efficiently generating, entangling, and detecting them is difficult. Other approaches (neutral atoms, quantum dots, topological qubits) present their own distinct trade-offs between stability, scalability, speed, and complexity, underscoring the diverse engineering obstacles on the path to fault-tolerant quantum computing.

Ultimately, solving complex, real-world problems requires quantum computers with thousands or millions of reliable qubits working in unison—a significant leap from current systems. Scaling up introduces formidable difficulties: controlling a larger number of qubits precisely becomes exponentially more complex, and the risk of errors propagating through the system increases. Maintaining qubit stability and coherence against decoherence becomes harder as systems grow, demanding advanced cryogenic shielding, error correction, and noise reduction techniques. Therefore, overcoming the intertwined challenges of scalability (increasing qubit count, connectivity, and control) and stability (maintaining coherence and managing errors via QEC, improved materials, and qubit design) is pivotal to realizing the full potential of quantum computing.

Common Misconceptions About Quantum Computing

As quantum computing captures growing attention, numerous misconceptions about its capabilities and potential impact have emerged. Addressing these is critical for a balanced understanding of what quantum computers can achieve and their inherent limitations.

1. Myth: Quantum computers will replace classical computers.

Reality: Quantum computers are specialized tools designed for specific problem types infeasible for classical systems (like complex simulations or certain optimizations). Classical computers remain superior for everyday tasks (word processing, web browsing, databases) due to their cost-effectiveness, efficiency, and mature ecosystem. Quantum computers are best viewed as complementary accelerators, augmenting classical systems in specific domains, not replacing them entirely.

2. Myth: Quantum computers are just much faster classical computers.

Reality: Quantum computing is a fundamentally different paradigm based on quantum mechanics. It leverages phenomena like superposition and entanglement to process information in ways classical computers cannot emulate, regardless of speed. Algorithms like Shor's (factoring) and Grover's (search) demonstrate speedups derived from these unique quantum principles, highlighting capabilities beyond mere classical speed increases.

3. Myth: Quantum computers will instantly break all encryption.

Reality: While algorithms like Shor's threaten current public-key cryptography (e.g., RSA), breaking today's standards requires large-scale, fault-tolerant quantum computers which are likely many years away. Furthermore, symmetric encryption (e.g., AES) is considered more resistant. The cybersecurity community is actively developing and standardizing Post-Quantum Cryptography (PQC) for a gradual transition, meaning an overnight cryptographic breakdown is highly unlikely.

4. Myth: Quantum computers can solve any problem quickly.

Reality: Quantum computers offer significant advantages only for specific problem classes where quantum algorithms provide a known speedup (e.g., factoring, quantum simulation, certain optimizations). For many computational tasks, they offer no known advantage and may even perform worse due to overheads and error rates. Classical algorithms remain essential for a vast range of applications.

5. Myth: Quantum computers are ready for widespread practical use now.

Reality: Despite rapid progress, quantum computing is still largely developmental. Current Noisy Intermediate-Scale Quantum (NISQ) devices have limited qubits and suffer from high error rates and short coherence times. Achieving "quantum advantage" on specific, often contrived tasks is a research milestone but doesn't equate to broad practical usefulness. Reliability, scalability, and effective error correction are needed for solving valuable real-world problems.

6. Myth: Building quantum computers is only about increasing qubit count.

Reality: While more qubits are necessary, progress hinges equally, if not more, on better qubits. Qubit quality (fidelity), coherence times, connectivity, and error management are critical. Simply adding noisy qubits can worsen performance. Effective quantum error correction (requiring many high-quality physical qubits per logical qubit) and advancements in control systems, materials, and integration are crucial for scalability.

7. Myth: Quantum computing's main benefit is speed.

Reality: While speedup for certain problems is significant, the true potential often lies in enabling entirely new capabilities. This includes accurately simulating quantum systems for breakthroughs in drug discovery and materials science, developing novel cryptographic methods like Quantum Key Distribution (QKD), and enabling new communication protocols for a future quantum internet. These represent new computational paradigms, not just faster versions of old ones.

Understanding these realities is essential for realistic expectations. Quantum computers promise to be powerful, specialized tools for tackling currently intractable problems, complementing classical computers within the broader computational landscape. Recognizing their unique strengths, limitations, and developmental stage allows for informed discussion about their transformative potential.

II. Opportunities, Risks, and Harms of Quantum Computing

Having established a foundational understanding of quantum computing principles and the challenges inherent in its development, the focus now shifts to the broader implications of this potentially transformative technology. Quantum computing promises unprecedented capabilities, but like any powerful innovation, its development and deployment are accompanied by significant considerations regarding its potential benefits, risks, and the ethical frameworks needed to guide its trajectory.

This exploration delves into the multifaceted landscape of quantum computing's impact. It examines the exciting opportunities across various sectors, from scientific discovery and healthcare to optimization and security. Concurrently, it provides a comprehensive analysis of the associated risks—spanning social equity, legal and governance structures, and safety concerns. Understanding these potential downsides is crucial for developing effective mitigation strategies and ethical guidelines, drawing lessons from historical technological advancements and considering approaches for future-proofing innovation through alignment with societal values. Navigating the complex ethical dilemmas presented requires careful thought and structured approaches to ensure a responsible and beneficial quantum future.

Opportunities in Quantum Computing

Quantum computing holds the potential for groundbreaking advancements across multiple domains, promising to solve complex problems currently beyond the reach of classical computers. Its unique capabilities open up transformative opportunities in science, technology, industry, and security, enabling solutions previously considered intractable.

One significant area of opportunity lies in enhancing cryptography and network security. Quantum Key Distribution (QKD), for instance, leverages quantum mechanics to establish ultra-secure communication channels where any eavesdropping attempt inherently reveals the intrusion, offering theoretically unbreakable encryption vital for sensitive data. Furthermore, the concept of a quantum internet promises not only fundamentally secure data transmission but also the ability to connect quantum computers for distributed processing power, tackling even larger collaborative problems.

In scientific research and healthcare, quantum computers offer revolutionary potential, primarily through their ability to simulate complex quantum systems—a task prohibitive for classical machines. This capability can drastically accelerate drug discovery and material science by accurately modeling molecular interactions, paving the way for new pharmaceuticals and materials with extraordinary properties, such as improved catalysts or components for energy solutions like batteries and solar cells. Understanding protein folding, crucial for biology, can be advanced by efficiently exploring possible configurations. This simulation power extends into healthcare, enabling truly personalized medicine by accelerating genetic analysis for tailored treatments and optimizing complex therapeutic strategies. Quantum simulation also offers profound opportunities in fundamental science, allowing researchers to probe the basic laws of physics in new ways.

Quantum computing is also poised to significantly impact optimization problems and artificial intelligence. Quantum algorithms provide powerful tools for tackling complex optimization challenges across industries, such as improving routing, scheduling, and inventory management in supply chains and logistics; optimizing financial portfolios; or enhancing the efficiency and stability of energy grids. In AI and machine learning, quantum approaches may improve the processing of high-dimensional data, potentially enhancing pattern recognition and predictive analytics, while new quantum-inspired architectures could lead to breakthroughs in solving complex AI problems.

Beyond these computational applications, quantum effects enable advances in measurement and sensing. Quantum metrology promises dramatic improvements in precision for navigation systems, timekeeping, medical imaging, and geological surveying, facilitating new scientific discoveries. These technological advancements are also expected to stimulate economic growth and workforce development, creating demand for new expertise in quantum science and engineering and driving innovation across the global market.

In summary, quantum computing presents a wide spectrum of opportunities—from revolutionizing materials science and medicine to securing communications and optimizing complex systems—promising to reshape industries and accelerate scientific progress by tackling problems previously considered unsolvable.

An Overview of Quantum Computing Risks and Harms

Quantum computing, while promising transformative advancements, also presents a complex array of interconnected risks spanning social equity, legal frameworks, governance structures, and overall safety. Proactive identification and mitigation of these risks—understanding both the potential for negative events (risks) and their damaging consequences (harms)—are crucial for responsible development and deployment.

Social Risks and Harms

A primary social risk stems from the immense cost and specialized infrastructure required for quantum computing, which risks creating a stark "quantum divide." This concentration of capabilities, primarily in more economically developed regions, could lead to harms such as exacerbated global inequalities, mirroring existing economic disparities (the "quantum gap"), and the loss of digital sovereignty for less-resourced nations.

Furthermore, the inherent complexity of quantum science presents the risk of "knowledge asymmetry." This gap hinders meaningful participation by non-experts, potentially harming democratic processes through reduced transparency and less effective governance. This challenge may be amplified by a "quantification bias," where focusing only on computable problems risks marginalizing essential social and ethical knowledge, leading to incomplete or biased decision-making.

Ethical discourse itself faces risks of bias and exclusion, often dominated by perspectives from certain regions. This poses the harm of "representational exclusion" for marginalized communities globally. Coupled with this is the need to address "intergenerational justice," as decisions made today pose the risk of negatively impacting future generations if long-term consequences are not considered.

The labor market faces the risk of disruption from automation driven by quantum capabilities, potentially harming individuals and communities through job displacement. Concurrently, the field confronts risks associated with a limited global talent pool, marked by gender imbalances and lack of diversity, which could harm innovation potential and equitable participation.

Finally, the enhanced data processing power of quantum computers introduces considerable risks to privacy and freedom of expression. The potential for pervasive surveillance by state or private actors could harm individual privacy on an unprecedented scale and create a chilling effect on open communication.

Legal and Governance Risks and Harms

A critical risk lies in the threat quantum computers pose to current cryptographic methods, particularly public-key algorithms. The potential to break widely used encryption creates risks of unprecedented data breaches, potentially harming individuals and organizations through compromised communications, financial losses, and significant legal liabilities. The "harvest now, decrypt later" scenario exacerbates this, posing a long-term risk to data confidentiality even before large-scale quantum computers are available, necessitating the development and deployment of quantum-resistant cryptography (PQC).

Compounding these technical threats is the risk stemming from inadequate governance frameworks. The current lack of comprehensive methods for assessing quantum's broad impacts, combined with the potential inadequacy of existing laws, risks allowing harms to emerge unchecked. This highlights the need for proactive ethics and potentially new regulations, avoiding the pitfalls of reactive approaches seen with past technologies.

Further legal uncertainty arises from the risk of unclear accountability and responsibility for quantum system failures or misuse, especially in critical sectors. This lack of clarity could harm victims by leaving them without recourse and hinder the adoption of beneficial technologies due to unresolved liability issues. Similarly, quantum access via cloud platforms introduces risks related to cross-border data governance, potentially leading to harms like jurisdictional conflicts or inequitable access.

Intellectual property rights face risks from the complexities of quantum innovation, potentially harming the balance between incentivizing inventors and ensuring broad access if patent eligibility, ownership, and enforcement are unclear or exploited. There's also a risk of market concentration and anti-competitive behavior, driven by high development costs and strategic actions like startup acquisition or aggressive patenting, which could harm the innovation ecosystem by stifling competition.

Lastly, exaggerated hype and speculation pose a risk by potentially misleading stakeholders. This could harm progress by distorting funding priorities away from fundamental research or safety, and by delaying necessary ethical oversight and regulation.

Safety-Related Risks and Harms

Quantum technologies present significant risks related to misuse and dual-use. Their power could be exploited for harmful purposes, such as developing advanced autonomous weapons, enabling mass surveillance, decrypting sensitive communications for espionage, or simulating dangerous biological/chemical agents. Such misuse could inflict direct harm through conflict, oppression, or biosecurity incidents. This necessitates strict oversight mechanisms.

The geopolitical risk of a destabilizing "quantum arms race," driven by the pursuit of quantum advantage for military superiority, is substantial. This competition could harm international relations by escalating tensions, eroding trust, increasing conflict potential, and diverting resources from other critical global needs. Robust international dialogue and cooperative security frameworks are needed to mitigate these dangers.

The inherent complexity and potential "black box" nature of quantum algorithms create risks related to opacity and explainability. This can harm public trust and accountability by making verification difficult and hindering the ability to anticipate or mitigate unintended negative outcomes from complex quantum systems. Furthermore, dependence on quantum technologies introduces risks; failures in critical systems could cause widespread societal disruption, while the quantum systems themselves could become targets for cyber-attacks, leading to cascading failures or data compromise.

Finally, there are environmental risks associated with the quantum technology lifecycle. The sourcing of rare materials, the high energy consumption of cryogenic systems, and the disposal of potentially hazardous components pose risks of environmental degradation and potential health hazards, requiring a focus on sustainable materials and practices.

This consolidated overview highlights the multifaceted nature of risks associated with quantum computing, emphasizing the need for integrated strategies across social, legal, governance, and safety domains to prevent potential harms.

III. A Practical Guide to Quantum Ethics

While the preceding discussions outline the foundational principles, potential opportunities, and significant risks associated with quantum computing, realizing its benefits while mitigating harms requires deliberate and proactive effort. Although these considerations affect society as a whole, they hold particular relevance for the scientists, engineers, and developers actively creating these powerful technologies. This necessitates robust ethical frameworks, thoughtful governance structures, and strategic foresight integrated directly into the innovation process.

This part delves into the crucial aspects of guiding quantum technology responsibly. It begins by detailing specific mitigation strategies designed to address the social, legal, governance, and safety risks previously identified. Drawing insights from historical precedents in technology ethics, it then explores actionable lessons for avoiding past mistakes. Furthermore, it examines strategies for future-proofing quantum innovation by aligning development with societal values and anticipating regulatory landscapes. Finally, it addresses the complex ethical dilemmas inherent in this field, proposing structured approaches to navigate challenging decisions. To make these considerations more tangible, particularly for those with technical backgrounds, subsequent discussions draw parallels between ethical analysis and familiar engineering concepts like quality control, aiming to bridge the gap between technical development and ethical responsibility and foster a quantum future that is both innovative and beneficial for humanity.

Mitigation Strategies for Quantum Computing Risks and Harms

Proactive ethical analysis serves as a vital quality control mechanism throughout the quantum computing development lifecycle. Analogous to quality assurance in traditional engineering, which aims to prevent defects and ensure reliability, ethical analysis works proactively to align quantum technologies with societal values and mitigate potential harms. This involves identifying potential negative impacts early and establishing clear guidelines for responsible development, thereby ensuring the "quality" of quantum technology regarding its broader societal consequences. Key practices include embedding ethical impact assessments, providing ethics training for developers, and establishing mechanisms for continuous review and adaptation.

Translating the awareness of potential risks—social, legal, governance, and safety-related—into actionable mitigation strategies is essential. Specific ethical guidelines and proactive measures can address potential failure modes, mirroring the rigorous testing and validation procedures used in engineering.

Mitigating social risks involves several approaches. Ethical guidelines should champion equitable access initiatives (e.g., cloud platforms, open-source projects, international collaborations like the Open Quantum Institute model) to counter the "quantum divide." Addressing knowledge asymmetry requires promoting public understanding via outreach and establishing participatory mechanisms for diverse input into governance. Ensuring inclusive discourse involves actively fostering diversity in ethical debates and using foresight processes that consider intergenerational equity. Countering potential labor impacts necessitates substantial investments in education, retraining, and just transition frameworks. Finally, mitigating privacy and surveillance risks demands emphasis on data minimization, transparency, clear limits on surveillance, and embedding privacy-by-design principles.

Addressing legal and governance risks requires managing encryption threats by prioritizing a timely transition to post-quantum cryptography (PQC) through research and standardization (e.g., NIST's work). Improving assessment involves mandating comprehensive impact assessments before deployment and fostering a culture of proactive ethical reflection. Establishing accountability necessitates developing clear legal and governance frameworks for liability regarding system failures or misuse. Navigating cross-border data governance calls for international agreements and standards for data sharing and cloud access. Addressing IP complexities requires ethical frameworks balancing innovation protection with widespread access, possibly via open innovation. Preventing market concentration involves advocating for fair competition and proactive regulatory assessment, while countering hype demands responsible communication standards and independent critical evaluation by policymakers.

Safety-related risks demand strategies such as robust oversight mechanisms (including potential ethical review boards and international dialogue) for misuse and dual-use potential (e.g., weapons, biosecurity). Mitigating cybersecurity threats requires emphasizing quantum-resistant measures and international cooperation. Addressing geopolitical instability involves promoting international dialogue, transparency, and collaboration to guide technology towards peaceful uses. Managing opacity and unintended consequences necessitates championing transparency, explainability, and rigorous testing. Lastly, addressing environmental impacts requires mandating full lifecycle assessments and promoting research into sustainable materials and energy-efficient designs.

By systematically applying ethical analysis and implementing corresponding guidelines addressing this comprehensive range of risks, the development of quantum computing can be steered towards maximizing benefits while diligently minimizing potential harms.

Applying Historical Lessons to Quantum Ethics

Examining historical failures in technological ethics provides critical, actionable lessons for navigating the complex ethical landscape of quantum computing. Understanding how ethical oversights led to past crises informs the proactive ethical governance needed for responsible quantum development. Past technological advancements unfortunately yield stark examples of such failures, including horrific disregard for human dignity and consent (e.g., unethical experimentation during WWII, the Tuskegee Syphilis Study), profound societal risks from large-scale data misuse for manipulation (e.g., the Facebook-Cambridge Analytica scandal), and severe harms from corporate deception and neglect of broader impacts (e.g., the Volkswagen emissions scandal).

These historical precedents offer direct insights relevant to the risks and mitigation strategies previously outlined for quantum computing. Specifically, the past disregard for human rights underscores the absolute necessity for robust ethical review mechanisms in all quantum research, mirroring the call for specialized oversight concerning dual-use research potential. It reinforces the importance of promoting inclusive discourse, ensuring diverse global voices and considerations of intergenerational justice are central to prevent repeating past exclusions. Similarly, historical data misuse amplifies the urgency of addressing quantum's potential impact on privacy and surveillance, strongly validating the ethical imperatives to implement privacy-by-design, ensure transparency, establish clear limits on surveillance, and promote robust cross-border data governance frameworks before quantum capabilities exacerbate vulnerabilities.

Furthermore, instances of corporate deception highlight the critical need for ethical conduct and transparency within the quantum industry. This connects directly to mitigation strategies such as countering hype through responsible communication, promoting fair competition to avoid market power abuses, addressing intellectual property complexities ethically, and undertaking comprehensive environmental impact assessments. It stresses that ethical considerations must permeate corporate culture, not merely function as compliance checklists. The overarching lesson is that neglecting ethics constitutes a significant risk—leading to severe reputational damage, legal penalties, financial losses, and profound societal harm. This reinforces the argument that ethical analysis is not an impediment but a crucial form of quality control and proactive risk management. Embedding ethical frameworks, impact assessments, and continuous review throughout the quantum development lifecycle is essential for building trust and steering the technology towards beneficial outcomes.

By internalizing these lessons, the quantum community can strive to avoid repeating past mistakes, fostering an ecosystem where innovation proceeds hand-in-hand with ethical foresight and a deep commitment to societal well-being, thereby building a more responsible and trustworthy quantum future.

Future-Proofing Quantum Innovation through Ethical Alignment

Ensuring the long-term success and societal benefit of quantum computing requires more than technical prowess; it demands ethical foresight—the practice of proactively aligning innovation with evolving societal values and anticipating regulatory trends. Technologies deeply rooted in ethical considerations are more likely to gain public trust, achieve widespread adoption, and prove sustainable over time. Integrating ethics from the outset is not merely a compliance exercise but a strategic imperative yielding tangible benefits: aligning with societal values fosters public confidence crucial for a "social license to operate," while organizations known for responsible innovation attract top talent and investment more readily. Proactively addressing ethical concerns helps prevent future legal challenges, public backlash, and costly interventions, especially given the regulatory trends seen with other emerging technologies like AI. Furthermore, proactive engagement with policymakers helps shape a supportive regulatory environment for responsible innovation.

Ethical considerations should be woven into the fabric of quantum technology development, mirroring core engineering principles. Embedding ethics early—just as engineers design for security and performance from the start—leads to more robust and trustworthy technologies. This ethical alignment ensures long-term societal acceptability and viability, much as maintainability ensures a system's continued function. As quantum technology and societal norms evolve rapidly, ethical frameworks must be scalable to handle increasing complexity and demonstrate ethical agility—adapting to shifting values, analogous to how crypto agility allows adaptation to new cryptographic standards. This necessitates continuous ethical monitoring and evaluation, mirroring engineering risk assessment, while designing with future ethical landscapes in mind embodies ethical forward compatibility. Engineering concepts like modularity and standardization can also inform the development of adaptable ethical frameworks across diverse quantum applications.

Viewing ethics proactively prevents the accumulation of "societal debt"—unresolved ethical issues that, like technical debt, can compound over time, hindering future progress. Ethical foresight acts as a "compass," guiding development towards beneficial outcomes aligned with human values and legal norms, a principle central to frameworks like Responsible Research and Innovation (RRI). Neglecting this guidance risks developing powerful technologies fundamentally misaligned with the societies they aim to serve. Ultimately, future-proofing quantum innovation hinges on recognizing that ethical considerations are integral to the technical and strategic development process. This holistic approach is essential for realizing the full positive potential of quantum computing in a sustainable and socially responsible manner.

Embedding Ethics for a Responsible Quantum Future

The development and deployment of quantum computing inevitably present complex ethical dilemmas where potential benefits clash with significant risks, requiring careful navigation. Applying a structured, risk-based ethical framework provides a valuable methodology for engineers, scientists, developers, and policymakers to make responsible choices when faced with these challenges. Ethical dilemmas frequently arise in areas previously discussed, such as the dual-use potential of quantum cryptography (balancing security breakthroughs against encryption-breaking threats), the societal impacts of job automation versus efficiency gains, and the equitable distribution of resource-intensive quantum capabilities versus their potential for groundbreaking discovery.

A risk-based ethical framework addresses such dilemmas not by providing simple answers, but by structuring the decision-making process. This involves first systematically identifying and evaluating both the potential benefits and harms associated with different development or deployment pathways. This is followed by assessing the likelihood and magnitude of these potential positive and negative impacts, considering various stakeholders and societal values like fairness, privacy, safety, and equity. Crucially, the process requires explicitly integrating the ethical principles and mitigation strategies outlined earlier (such as promoting equitable access, ensuring privacy-by-design, transitioning to PQC, fostering transparency, supporting workforce transitions, etc.) into the evaluation. This culminates in making informed and justifiable decisions about trade-offs, prioritizing actions that maximize overall benefit while diligently minimizing harm, particularly to vulnerable groups. This structured approach, analogous to technical risk assessment in engineering, moves beyond intuition, allowing for a more objective, transparent, and ethically grounded way to navigate the complex trade-offs inherent in developing powerful new technologies like quantum computing.

Integrating ethical considerations throughout the lifecycle of quantum computing is a fundamental necessity for responsible innovation. As established in previous sections, ethical foresight and proactive alignment function as essential components of effective risk management, robust quality control, and strategic future-proofing for this transformative field. Realizing the immense potential of quantum computing for societal good requires a sustained, collaborative, and proactive effort. Engineers, scientists, industry leaders, policymakers, ethicists, and the public must work together to shape norms, guidelines, and regulations that ensure quantum technologies are developed and deployed safely, equitably, and in alignment with human values. Ultimately, the success and lasting positive impact of the quantum era will depend not only on technical breakthroughs but equally on our collective commitment to navigating its ethical dimensions with wisdom, responsibility, and a steadfast focus on the well-being of humanity.

Subscribe now

AI models such as ChatGPT, Gemini, Claude, DeepSeek, and Copilot have quickly evolved beyond simple tools for chatting, retrieving information, and editing text. Increasingly, they serve as powerful assistants for complex tasks, including advanced mathematical analysis, large-scale data processing, and programming. Their rapid development raises an important question: how is it possible for AI to learn and advance so quickly?

How Do Humans Learn?

To understand how AI learns, it is useful first to examine how humans acquire and deepen their knowledge. When approaching a new subject, we typically begin by studying literature and other reliable sources of knowledge, which help us grasp key concepts and build a fundamental understanding of the field. The quality and accessibility of these resources significantly influence the speed and effectiveness of our learning, as well-structured material is easier to integrate into a broader mental framework. The goal of this initial phase is to establish a strong theoretical foundation that allows for further development and successful application of knowledge in practice.

Once we have a solid grasp of the basics, we move on to solving specific problems that systematically guide us through problem-solving processes. In this phase, we follow established methods provided by teachers or textbooks, enabling us to develop analytical skills and learn the correct approaches to tackling challenges. Additionally, as we encounter various tasks, we adapt our strategies, enhancing our ability to think critically and creatively. This transition from theory to practice is crucial, as it ensures that knowledge is not merely abstract but becomes a dynamic tool for solving real-world problems in diverse situations.

The third stage of learning involves independent problem-solving, where we are given only the final result without a step-by-step solution. This requires us to develop appropriate strategies and find optimal approaches beyond merely repeating previously learned methods. Some problems may not be solvable using conventional approaches, necessitating the creation of innovative and more effective solutions. In such cases, creative thinking, adaptability, and the ability to experiment with different solutions are essential. Additionally, by verifying the correctness of our methods against known outcomes, we refine our problem-solving approaches, gaining deeper insights and developing adaptable strategies for tackling even more complex challenges.

Three Stages of AI Learning

These three fundamental and interconnected learning stages, essential in human learning, also form the basis for training neural networks. Just as humans gradually build knowledge through theoretical understanding, practical application, and independent problem-solving, neural networks follow a similar process.

1. Foundational Learning (Pretraining)

In the first stage of AI learning, large language models analyze vast amounts of carefully curated and structured data, including scientific papers, literary works, journalistic articles, online discussions, and forum posts. This phase, known as foundational learning (pretraining), functions like extensive "independent reading": the model learns to recognize and understand patterns and relationships within data, much like a student who first thoroughly studies theory from textbooks.

This learning process often relies on self-supervised learning approaches, where the model gradually develops the ability to predict the next word or, more precisely, the next "token" in a text. During this phase, the model constructs a rich internal representation of linguistic structures, concepts, and styles, allowing it to generate coherent, grammatically correct, and stylistically appropriate text. Based on an initial sentence, it can "predict" the most likely continuation of the text by drawing on contextual patterns observed in vast datasets.

However, at this stage, foundational AI models primarily function as statistical predictors of language patterns rather than as reliable sources of factual information. Their responses are often well-formed stylistically but not necessarily accurate or meaningful. To enhance their precision and relevance, further training is required.

2. Task-Specific Learning (Fine-Tuning)

Although models learn language comprehension and general pattern recognition in the first phase, theoretical knowledge alone is often insufficient for solving specific tasks. After foundational training, neural networks refine their broad theoretical knowledge by learning how to apply it practically in conversations and problem-solving. This phase, known as fine-tuning, involves training on large sets of practical examples where questions or prompts are paired with verified correct answers.

Similar to learning mathematics, where we first master theory and then practice with concrete problems, neural networks in this second phase gain experience by studying previously solved problems and answered questions. This enables them to learn the correct approach to solving tasks and formulating responses.

Fine-tuning requires significantly less time than foundational learning since its primary focus is on optimizing responses for specific applications. The result is AI models that can effectively communicate and answer questions. However, their responses may still be unreliable when dealing with topics not well covered in their training data.

3. Independent Problem-Solving (Reinforcement Learning)

In the third learning phase, neural networks actively experiment with different approaches to problem-solving and continuously optimize their solutions. This method allows them not only to reinforce existing patterns but also to develop entirely new strategies that may surpass human intuition and approaches. The role of feedback (e.g., correct solution of tasks or "rewards" and "punishments") is crucial in this process, as it serves as a measure of success for the model and helps it identify the most effective tactics. This method, known as reinforcement learning, enables AI to refine its decision-making based on iterative feedback.

A key advantage of this approach is the ability to adapt to unforeseen situations and challenges not present in the training data. Instead of relying solely on a static set of correct answers, the model autonomously explores and tests new possibilities. This creates a dynamic learning process where the model adjusts its hypotheses and selects optimal strategies based on its environment and received feedback.

Through independent problem-solving, neural networks move beyond mere replication of learned patterns and develop highly flexible, creative, and contextually relevant problem-solving methods. This advanced learning stage is essential for tackling complex challenges across various domains. By combining exploratory learning with existing knowledge, neural networks not only refine known solutions but also discover new ones. Thus, they can develop entirely original problem-solving approaches that differ from the methods they were trained on.

AlphaGo and Move 37

The first major breakthrough of this kind of "innovative" learning occurred in 2016, during the legendary Go tournament, when the artificial intelligence AlphaGo from the company DeepMind competed against the top Korean player Lee Sedol. In the second game, AlphaGo executed the now-famous move 37, which astonished experts and Go enthusiasts worldwide. The move was so unconventional and outside established human strategies that it was initially dismissed as a mistake. However, it later proved to be a brilliant play, placing Lee Sedol in a difficult position from which he could not recover. Visibly shaken, he admitted to underestimating AI's ability to surpass human strategies.

AlphaGo did not derive move 37 from analysing past games but developed it through extensive self-play, where it experimented with different strategies and identified the most successful ones based on feedback. By playing millions of games against itself, the model refined entirely new tactics unknown to human players. Move 37 marked a turning point in AI development, demonstrating that systematic independent learning could enable AI to surpass human capabilities.

Just as AlphaGo once proved that a neural network could master a strategic game beyond human expertise, today, large language models use the same methods to improve their abilities in solving complex problems. The third phase of AI learning, involving independent problem-solving through reinforcement learning, has been particularly refined and optimized by researchers in China during the development of the DeepSeek-R1 model.

Subscribe now

During World War II, the American intelligence service published a manual for civilians in occupied territories, encouraging them to engage in small but effective acts of sabotage. The instructions in the Simple Sabotage Field Manual were surprisingly straightforward and accessible to anyone.

Among other things, the manual advised potential saboteurs to weaken constructive discussion at various levels of society. They were instructed to prolong meetings with endless debates over trivial details, enforce every regulation to the letter, spread false or contradictory information, and otherwise sow confusion. At first glance, these tactics seemed trivial and posed minimal risk to the perpetrators, yet they could significantly disrupt the enemy's military, industrial, and administrative systems.

Algorithms and the New Age of Polarization

Today, in the digital age, it seems we have unintentionally become victims of a similar form of sabotage to constructive public discourse. In an increasingly polarized society, even reaching agreement on simple, easily verifiable facts has become a challenge. However, this time, the culprits are not hostile foreign agents but rather the very technological tools we rely on to make our work and communication more efficient.

Social media algorithms are designed to maximize user engagement by showing content that captures our attention. This approach has benefits, such as faster access to relevant information, like news about scientific breakthroughs that interest us. However, these algorithms are not tuned to serve the broader public good, which significantly impacts the quality of public debate.

The dark side of algorithm-driven content recommendation is its role in fostering societal polarization. Algorithms often amplify content that provokes strong emotions, such as outrage, mockery, or anger. While such triggers can be valuable—for instance, in exposing misconduct—they become problematic when they serve political propaganda or deepen societal divides.

Democracy in the Crossfire of Misinformation

Democracy is built on the premise that citizens make decisions based on verified information and a shared understanding of reality. In the digital age, this assumption is increasingly under threat as traditional media, which provided fact-checking and editorial standards, lose influence. The informational landscape is now dominated by viral content, where quality often takes a backseat to sensationalism.

This shift is not just a change in how information is distributed but a fundamental transformation in how social reality is constructed. Professional journalists and editors, who once acted as gatekeepers of information, are increasingly replaced by influencers and algorithms. These new forces prioritize content that garners the most attention, not necessarily what is true or socially significant. The result is a fragmentation of reality, where different groups live in informational bubbles that reinforce their beliefs, deepening polarization and eroding trust in institutions.

Can Democracy Survive in a Post-Truth Era?

Without a shared understanding of basic facts, the foundations of democratic discourse crumble. Dialogue and collaboration give way to conflict between opposing "truths," making collective decision-making nearly impossible. Democracy, always a fragile system, now faces the threat of being undermined from within by its own informational infrastructure. This raises a critical question: How can we rebuild a shared reality in the age of digital platforms and algorithms—a reality upon which trust, dialogue, and democratic decision-making depend?

The paradox is that we live in an era with unprecedented access to information, yet achieving societal consensus on basic facts is becoming increasingly difficult. While an abundance of information might seem to strengthen democracy by enabling more informed decision-making, the reality is often the opposite. The fragmentation of information sources, the flood of false or misleading content, and algorithms that reward sensationalism create an environment where realities among different groups not only diverge but directly contradict each other.

Rebuilding Trust in a Divided Information Landscape

The responsibility for maintaining the quality of public discourse does not rest solely on platforms and their algorithms but also on us as users. Mocking the "other side" by highlighting the most absurd or provocative claims, paired with ironic commentary, may be tempting and algorithmically rewarded, but it ultimately deepens polarization. We must resist this temptation and strive to share reasoned, verified information in ways that algorithms can recognize and amplify. This is not an easy task, but it is necessary. It is not just a technical challenge—it is an ethical imperative.

Democracy as a process requires dialogue, understanding, and at least a minimal level of agreement on what is true and what is not. Without this foundation, political decisions become mere clashes between irreconcilable "truths." If we cannot agree on basic facts, how can we, as a society, tackle complex challenges that demand collective action? Can democracy survive in a fragmented information space where truth is no longer universal but relative? Solving this problem is one of the defining challenges of our time.

Subscribe now

Over the past few years, the field of artificial intelligence (AI) has made remarkable progress. The advancements in the capabilities of AI tools are truly astonishing. Today, we can communicate with computers using our everyday language, and machines have mastered complex tasks that were once believed to be the exclusive domain of humans. The progress in improving the quality of smart devices has surprised even veteran experts in the field.

The essence of the AI revolution

AI technological revolution can be compared to the emergence of the internet a few decades ago, but at that time technological innovations such as email, remote file access and the World Wide Web were introduced much more slowly into everyday life. The power and utility of new AI tools such as chatGPT and various image generation systems is truly fascinating, and their easy accessibility has quickly led to mass adoption.

With the new smart language systems, we can talk to articles and books, ask them all sorts of questions, seek further clarification, and do many things we could not do with texts before. We can also conduct market analysis, draft a wide range of communications, reply to emails, summarise the content of meetings and many other similar time-consuming tasks. Nor is it a barrier to new technologies if the text we want to read is only available in a language we do not understand. The smart tools make it easy to discuss its content in English, even if the source text is in Arabic, Chinese or any of the many languages already supported by the new systems.

The proliferation of languages already covered by new AI tools has made human knowledge even more accessible to people around the globe. Researchers are now working to develop language models that support all the world’s languages for which sufficient digitised resources are available as a prerequisite for the implementation of machine learning.

What scientists don’t yet understand about AI

These technological advances are accompanied by claims that even scientists developing new AI tools do not have a good understanding of why there is such a leap in the capabilities and usefulness of this type of technology right now. Researchers are, of course, very familiar with the workings of the devices and programs that power smart tools. They also understand that in addition to understanding the theory, to effectively perform machine learning of AI models, they need a lot of computing power, which is not cheap. However, it is quite another thing for experts to be able to explain how AI actually performs the tasks that we believe require some kind of a thought process.

The frequent remark that large language models such as chatGPT can only statistically predict the most likely next word in sequence ignores the very essence of the revolution we are witnessing in the field of artificial intelligence. By learning to predict letters or words in a text, the neural network can also be taught many other tasks. 

Large neural networks can extract patterns from the mass of data they learn from, allowing them to perform many complex tasks, which is a remarkable achievement. In a way, this is similar to the phenomenon where systematic training in long-distance running would prepare an athlete to be equally successful in other sports. The result is strange and perhaps surprising, but it is an important element of the recent AI revolution.

How do machines perform tasks that require thinking?

Large language models of artificial intelligence, capable of running smart services such as chatGPT, are based on large neural networks. These are huge mathematical equations where, during the learning process, parameters are gradually adjusted so that the models are then able to calculate a meaningful answer to a given question. However, as with the connections between neurons in the human brain, it is not immediately clear how the structure of the connections between neurons translates into the calculated answers to questions in artificial neural networks.

Large-scale AI models operate on a black box principle, where input data passes through the complex connections of a large neural network, which can include billions of parameters or weights, but the way these parameters integrate with each other to produce specific results is not obvious. Understanding the “black box” problem, which concerns the performance of large-scale AI language models and functioning of the human brain, is one of the great scientific puzzles of our time. This challenge is not only important for building and managing even more powerful AI models in the future, but also represents one of the fundamental questions in AI research.

The key problem is not, of course, a lack of understanding of the mathematical principles that underpin how AI models work. The central problem lies in a deeper question: how is it possible that huge mathematical equations can so efficiently perform tasks such as answering questions, generating text, translating between languages, creating images and other similar activities that, until recently, only humans could do well?

The phenomenon of generalisation in AI

A fundamental element of machine learning is the phenomenon of generalisation. It represents the basic way AI models can learn to “understand” something, not just learn it by heart. Generalisation in machine learning is the ability of a model to efficiently and correctly predict or explain new, previously unknown data that comes from the same general population as the training data. In essence, it is the capacity of the model to apply the learnt knowledge from the training set to data that it has not seen during training, which is crucial for its practical applicability.

Models can learn to perform tasks such as translating sentences from one language into another by training on a set of already translated examples. However, they can generalise their knowledge and learn to perform similar tasks on examples they have not seen before. Models not only remember patterns they have already seen, but also independently develop rules during the learning process that enable them to apply these patterns to new examples. In particular, large language models such as GPT-4 have a surprising capacity for generalisation.

When we train an AI model, we want it to learn patterns that are generally valid for the problem we are trying to solve, but we do not want it to overfit to the specifics of the training dataset. If we overfit the model to the training data, it will perform excellently on the training data, but its performance on the new data will be much worse, because it has learned the specific details of the training set rather than developing an “understanding” of the general patterns.

The invention of the artificial neuron and training algorithm

The remarkable technological revolution in artificial intelligence that we have witnessed in recent years is the result of years of research, which has recently reached an important peak. Scientists have long wondered what makes the human brain intelligent, and many years ago concluded that the ability to think is most likely related to the number of neurons and how they are organised.

That is why, in the middle of the 20th century, researchers began to investigate how the workings of the neurons in the brain could be mimicked by mathematical models. This approach led to the invention of the artificial neuron, which is nothing more than a mathematical formula that attempts to emulate the functioning of a biological neuron. Many interconnected artificial neurons then form a neural network, which is also just a mathematical equation with many parameters.

The next major step in developing artificial neural networks was the discovery of a method by which artificial neurons can be trained. The backpropagation algorithm allows neural networks to learn from data or tasks that have already been solved. Simply put, the neural network calculates its prediction of the outcome and compares it to the actual outcome of the task, then calculates how much each neuron contributed to the error. In the next step, it corrects the parameter settings of each neuron so that the overall prediction error is reduced. If the process is repeated many times, the neural network parameter settings are gradually changed in such a way that the neural network becomes better and better at predicting the correct results.

The unsupervised learning process

Supervised machine learning means that we know in advance what we want the neural network to learn. For such learning, we need a large number of solved tasks on which to train the neural network. But unsupervised machine learning is a much more interesting approach, because we train the neural network to be as good as possible at a particular task, such as predicting the next word in a text, while at the same time it learns many other tasks that we have not trained it to do.

The unsupervised learning process of neural networks can also be seen as a process of discovering hidden structures in the data. We do not instruct a machine what to learn, but by teaching it to predict the next word, we enable it to systematically “read” texts and learn their content.

In 2017, OpenAI researchers taught a neural network to predict the next letter in a collection of 82 million reviews written by customers on Amazon for different products. An important result of this machine learning was that by training the neural network to predict the next character in texts, it learned not only this skill, but also a number of other skills for which it was not directly trained.

For example, they found that one of the neurons in the neural network became particularly sensitive to the mood of a particular text. In most cases, if the neuron was active, the rating was positive, but if it was inactive, the rating was negative. They also found that by switching this neuron on and off, they could directly control the sentiment of the newly generated reviews.

The experiment showed that learning to predict letters or words in a text can also teach the neural network many other tasks over time. In a way, this is similar to the phenomenon that systematic training in long-distance running would train an athlete for other sports. The result is unusual and perhaps surprising, but it is an important element of the recent AI revolution.

Knowledge as the  by-product of guessing

Over the next few years, when much larger neural networks were trained on a very large set of texts, it turned out that although the neural network only learns to predict the next word in a sequence of words, it somehow learns to “understand” the content of those texts. Of course, this “understanding” is not the same as in the human brain, but even in an artificial neural network, structures are formed that somehow correspond to the ideas, or a very condensed form of the notation, of the mass of information contained in the texts.

If the learning process is performed correctly, the neural network automatically extracts key ideas from the dataset during the learning process. These ideas are then stored in the neural network, which is a rather compressed record given the large amount of source data. The trick is that the neural networks we use for generative models have a much smaller number of parameters than the amount of data we train them on, so the models need to discover and effectively internalise the essence of the data to be able to regenerate it. By compressing it into a shorter record, they can extract key patterns from the data that enable understanding and prediction, which is certainly an impressive achievement.

The frequent observation that large language models such as chatGPT can only statistically predict the most likely next word in a sequence misses the point of the revolution we are witnessing in the field of artificial intelligence. Learning to predict the next word should be understood as a reading process, the by-product of which is the “knowledge” of the content that the neural network is reading.

Data compression and concept generation

The essence of unsupervised learning, where we optimise a neural network for one task while it learns something else that we are actually interested in, can also be thought of as a form of data compression. The complexity of a dataset can be defined as the shortest possible instruction to reproduce that data. While the machine is learning to predict the next word in the text, it manages to significantly reduce the size of the recording of the essence of this data, by managing to store it in the parameter settings or weights of the neural network.

Through generalisation and other similar processes, large AI language models perform a kind of compression of the knowledge they learn from. They condense large amounts of available information through generalisation into a well-structured compact form that takes up much less disk space than the original data. Although it is just a matter of setting parameters in a huge mathematical equation, this process of machine learning can also be seen as a kind of formation of concepts that otherwise form the basis of thought.

Although during the learning process the neural network only tries to correctly predict the next word in the large corpus of texts we are training it with, it also creates a kind of conceptual world that remains hidden in its parameters or weights after the learning process is complete. Artificial intelligence does not learn “by memorising”, but rather generalises and organises data by developing slightly differently constructed “concepts”. In doing so, it creates structures in the hidden (latent) mathematical space of the neural network. It could also be described as a kind of virtual world of ideas, as Plato imagined long ago when he tried to understand how people think and learn.

The next frontier for artificial intelligence

Artificial intelligence currently learns from textual data that contains ideas already organised and articulated through concepts. The conceptualisation, carried out by humans, is thus pre-established. AI synthesises and structures the pre-existing conceptual frameworks found in its training corpus into the mathematical latent space of a neural network.

The next significant breakthrough for AI will be its ability to conceptually structure and thus understand information that has not yet been conceptually processed or organised. This involves deriving insights from raw, unprocessed “sensory” data. This capability is partially demonstrated by AI systems powering autonomous vehicles, which interpret the road vehicle surroundings to ensure safe navigation and decision-making in traffic. However, even in this scenario, the categorisation of data is predetermined.

The potential for AI to devise different, novel, and alternative categorisations of the world opens up a vast array of intriguing and significant philosophical questions and dilemmas. The advancement towards AI that can independently conceptualise and categorise raw data without human intervention will mark a profound shift in our understanding of cognition and the nature of intelligence itself. This evolution challenges us to reconsider the boundaries of AI capabilities and its role in expanding the horizons of human knowledge and perception.

Summary: The article provides a comprehensive overview of the latest developments in artificial intelligence (AI), framing the AI revolution as a transformative period comparable to the advent of the internet. It delves into the complexities and challenges that researchers encounter in dissecting the mechanisms through which AI executes cognitively demanding tasks. Key topics covered include the concept of generalisation within AI systems, the backpropagation algorithm, and the dynamics of unsupervised learning. The article further explores how large-scale AI models manage to “understand” content through a process of data compression and concept generation, suggesting that AI models create a conceptual world within its mathematical space. The discussion culminates in an examination of the forthcoming horizon for AI: the capability to autonomously conceptualise and categorise unprocessed data, independent of human guidance.

This article was originally published in Bančni vestnik, Volume 73, No. 5, May 2024.

Subscribe now