Philosopher Dieter Mersch has a clear answer to the recurring claim that AI is racing toward superintelligence or consciousness. In a recent interview, he calls current systems mathematics machines. They optimize functions, recognize patterns, and maximize probabilities. The results can be striking. The underlying process is not mysterious.
The Mathematical Limits
Mersch draws on three classic limits from the philosophy of mathematics. Gödel’s incompleteness theorems show that no sufficiently complex formal system can fully describe itself. Turing argued that inventing new concepts requires intuition that sits outside pure formalization. Tarski demonstrated that formal systems generate only formal semantics; they do not produce genuine judgments of truth, relevance, or meaning. What cannot be mathematized remains inaccessible to them.
This is why self-correction and multi-agent setups do not change the fundamental picture. Those mechanisms remain recursive calculations. They improve outputs inside the existing mathematical structure. They do not step outside it.
Humans, by contrast, notice that the act of thinking always exceeds the content of any particular thought. That gap is not a bug. It is the condition for genuine reflexivity, creativity, and freedom. AI can simulate the appearance of reflection. It does not perform it.
Art, Imitation, and the Reflexive Stance
The same limit applies to art. Systems can imitate styles with high fidelity. They do not produce the kind of practice that questions its own seeing, its traditions, or the relation between image and world. That requires a reflexive stance that the mathematics does not supply.
The Cultural Risk
Mersch’s deeper warning is cultural rather than technical. The real risk is not that machines will replace human intelligence. It is that we will lower our standards until the two appear interchangeable. When student texts become smoother yet less distinctive, when journalism and research settle for competent surfaces, and when novelty is measured only by statistical surprise, culture flattens. We lose the awkward, specific voices that still carry actual thinking.
Connections to the Broader Debate
This view aligns with my practical stance already developed on this website. Treating AI as an oracle invites people to outsource judgment in medicine, finance, and everyday decisions. The ethical task is not to reject the tool but to keep human agency intact. The same point appears when AI is framed as a tool rather than a creature or a god.
Once the existential packaging drops, clearer questions become possible: what does this system actually do, where does it amplify existing strengths or weaknesses, and how should we design around it?
The need for a Mindful Revolution follows directly. Complex systems accumulate entropy. Nations, institutions, and technologies grow more intricate until ordinary corrections no longer restore order. Humans can still reflect on their own generative models and update them deliberately. That capacity is not automatic. It is trained through attention, self-observation, and the willingness to revise habits of thought.
AI lacks the equivalent. It cannot notice that it is operating inside a particular formal frame and choose to leave it. The human practice of observing both the technology and the observer therefore remains the scarce resource.
Software, Hardware, and the Boundary
Ralf Otte’s book Maschinenbewusstseinraises a related but distinct question. He argues that software running on conventional von Neumann architectures remains pure mathematics and cannot generate consciousness. Only a shift toward brain-inspired, neuromorphic hardware might produce something he calls machine consciousness — an elementary form tied to physical embodiment rather than human experience. Even that prospect, however, would still require a societal decision about how far to go.
Mersch’s position is stricter: as long as the system stays within mathematical formalization on conventional computer architectures, feelings and consciousness remain outside its reach. Simulating those phenomena does not produce them.
What Remains Distinctly Human
The practical consequence is straightforward. AI can accelerate pattern recognition, code generation, and certain forms of optimization. It does not supply the reflexive judgment, the sense of meaning, or the capacity to step outside a given frame that human thinking still provides.
Keeping that distinction clear protects both the usefulness of the tools and the human capacities that make wise use possible. The alternative is a gradual substitution of surfaces for substance — one that becomes visible only after the flattening is already widespread.