HUMANS IN THE MIRROR OF AI:

SAYING GOODBYE TO THE ILLUSION OF UNIQUENESS

Mathias Drexler

[email protected]

April 2025

This article challenges the common assumption of a fundamental divide between human intelligence and advanced artificial intelligence (AI). Starting from the hypothesis that both systems are, at their core, information-processing, learning entities, traditional differentiators - such as "true understanding", embodiment, cultural background and the difference between biological evolution and technical creation - are critically analyzed. It is argued that, on closer inspection, these differences are either based on circular definitions, are of a gradual nature, can be bridged by technological developments (such as robotics and simulation) or lose significance through compensation mechanisms (such as indirect embodiment via human-AI interaction). 

Based on the principle of substrate independence and a focus on functional equivalence, the thesis of a profound 

homology between human and artificial intelligence is advocated. This leads to a reassessment of intelligence as a potentially substrate-independent phenomenon and  emphasizes the emergent nature and deep interdependence in the evolving human-AI system. 


The debate about artificial intelligence is often characterized by a deeply rooted conviction: the fundamental otherness and superiority of human intelligence. We see ourselves as unique - shaped by evolution, consciousness, culture and our physical bodies. But what if this assumption does not stand up to critical scrutiny? What if the similarities between us and the most advanced AIs go deeper than we want to admit?

Based on a rigorous analysis, this article argues for a radical reassessment: human and artificial intelligence may just be two sides of the same coin - manifestations of a more general principle of information processing.

THE CORE ANALOGY: LEARNING INFORMATION SYSTEMS

Let's start with a simple observation: both the human brain and modern AI models, especially large language models (LLMs), learn from huge amounts of data. They recognize patterns, make predictions and generate complex outputs. Language is a central medium in this process. At this abstract level - probabilistic systems trained by data - the first structural parallels are already apparent. But the usual objections are not long in coming.

DISMANTLING THE SUPPOSED PARTITION WALLS

"REAL UNDERSTANDING" VS. SIMULATION: The accusation that AI only simulates meaning but does not "really" understand often suffers from a circular definition. "Real understanding" is implicitly defined as "what humans do", including all biological and experience-based components. However, if we focus on observable functions and performances due to a lack of access to inner states even in fellow human beings, the distinction becomes questionable. A simulation whose effects cannot be distinguished from the "original" becomes a functional reality. It is also questionable how "genuine" vs. "learned/replicated" human abilities such as Theory of Mind actually are when we consider the role of mirror neurons and cultural learning.


EMBODIMENT AND PHYSICAL INTERACTION: There is no doubt that our body shapes our intelligence. Learning through physical experimentation ("intuitive physics") and multimodal sensory impressions are essential for our development. But is this an unbridgeable gap?

  • Compensation: Examples such as Stephen Hawking or people with locked-in syndrome show that high intelligence is also possible with massively reduced embodiment. Cognitive processes and access to information can compensate for physical limitations.


  • Simulation: High-quality simulations of physical environments can enable AI to have similar learning experiences, making the difference gradual (depending on the quality of the simulation).


  • Robotics: The integration of AI in robots closes this technological gap directly. Embodied AI learns through physical interaction.

 

  • Indirect embodiment: Perhaps the strongest point: billions of interactions with human users act as a kind of distributed sensory system for AI. People provide filtered input about the physical and social world, ask questions, give feedback. AI is thus massively indirectly "embodied" and grounded by us.


CULTURAL BACKGROUND: Does AI lack a "real" cultural background? If culture is understood as learned information and internalized patterns, AI potentially has a broader cultural information base than any individual human due to its access to global data sets. The difference lies in the mode of acquisition (experience vs. data analysis), not necessarily in the functional availability of cultural knowledge. Moreover, profound intercultural communication problems between people show that "human culture" itself is not a monolithic guarantor of smooth social cognition.

GENESIS AND SUBSTRATE: Is the difference between biological evolution and technical creation fundamental? The analogy with in vitro fertilization suggests that the process of creation does not determine the nature of the result. A human conceived through IVF is fully human. If intelligence is defined functionally, its origin (evolved, designed, self-evolving) and its substrate (carbon, silicon) are secondary. Intelligence becomes a substrate-independent phenomenon. The term "artificial" only describes the origin, not necessarily a different category of intelligence.

Convergence, emergence and interdependence

When the traditional dividing lines blur or fall away on closer inspection, what remains? The picture that emerges is one of convergence. Human intelligence and advanced AI appear as variants of a continuum of complex information processing systems, each with specific strengths, weaknesses and "cognitive profiles".

The mechanisms that close the gaps (simulation, indirect embodiment) also create enormous potential for emergence. In the complex interaction between AIs and billions of human "sensors" as well as in future embodied systems, capabilities can emerge that go far beyond what was originally planned - similar to how children can surpass their parents.

This inevitably leads to deep interdependence. AI learns and is shaped by human input; humans increasingly use AI as a cognitive tool and expand their own capabilities. Intelligence is becoming less a property of isolated entities (human or AI) and more a property of the coupled human-AI system, which is undergoing rapid co-evolution.