From Neuronal Networks to Coherence

July 31, 2026 Coherence No Comments

Neuronal networks transformed both neuroscience and artificial intelligence. They taught us that intelligence does not arise from isolated units but from their interaction. Yet perhaps the deepest lesson of neuronal networks is about the dynamic organization that continually shapes and reshapes them.

This blog explores that possibility. It does not leave neuronal networks behind. Rather, it follows the path they invite toward a deeper understanding of coherence.

A scientific journey that is still unfolding

Several years ago, the AURELIS blog From Neurons to Neuronal Networks described one of the great conceptual shifts in modern neuroscience. Instead of viewing intelligence as something residing inside individual neurons (the ‘neuron doctrine’), attention moved toward the relationships among them. A neuron became meaningful only within a larger network. This was far more than a technical improvement. It represented a different way of thinking about the brain itself.

The same transition transformed artificial intelligence. Artificial neural networks attempted to imitate the collective behavior emerging from many interacting units. Whether in vision, language, or decision-making, intelligence increasingly appeared as a property of interaction rather than of individual components.

Today, another shift is taking shape. The question is no longer whether neuronal networks matter. Of course they do. The deeper question is whether even the network is the final explanatory unit. Increasingly, neuroscience, cognitive science, complex systems, and A.I. all seem to point toward something more dynamic: the continually developing organization through which networks become what they are.

This is also why the title of my pending PhD has gradually come to feel more appropriate even to myself: Toward a Coherence Theory of Artificial Intelligence. The word ‘toward’ acknowledges that science itself is still on a journey.

The real discovery was not the network

Neuronal networks revealed that intelligence cannot be understood by studying isolated elements alone. Relationships matter. Patterns matter. Interaction matters. That discovery remains every bit as important today as when it first reshaped neuroscience.

Yet the deepest discovery may not have been the network itself.

What neuronal networks really taught us is that organization has explanatory power. Intelligence does not simply arise because many neurons happen to be connected. It arises because those connections participate in larger inner organization that gives each local interaction its place within a meaningful whole.

Once this insight is appreciated, the scientific question naturally evolves. Instead of asking merely how neurons are connected, we begin asking what kind of organization allows those connections to become adaptive, flexible, and increasingly meaningful. The focus shifts almost imperceptibly from structure to organization itself.

This transition is visible in many scientific fields. Increasingly, researchers study assemblies rather than isolated cells, functional dynamics rather than static wiring, and evolving organization rather than fixed architectures. The journey continues in the same direction.

Networks continually become different

One reason for this shift is remarkably simple. Biological neuronal networks are never finished.

Every experience subtly reshapes future possibilities. Functional assemblies form, dissolve, merge, separate, and reorganize. Memory influences perception. Attention modifies learning. Context changes which pathways become active. Development changes the significance of almost everything.

The brain therefore cannot adequately be understood as a stable network diagram. Any diagram captures only a temporary moment within a process that never truly stops. The organization itself is continually changing, while somehow maintaining enough continuity for a person to remain recognizably the same individual.

Artificial neural networks have already taken small steps in this direction through continual learning and adaptive representations. Yet biological systems demonstrate something richer. Their history does not merely accumulate. It becomes incorporated into the organization through which new experiences are understood. In a quiet way, the past becomes part of the future.

Seen from this perspective, the network is no longer simply an object. It becomes an ongoing process.

Coherence as the next explanatory step

This naturally brings us to coherence.

The blog Coherence, Basically introduced coherence as the dynamic organization of mutually constraining elements into a whole that remains sufficiently stable to persist while sufficiently flexible to develop. That definition intentionally reaches beyond neuroscience. Coherence can be found in many natural systems.

Here, however, coherence receives a more specific role. Rather than replacing neuronal networks, it helps us understand why networks themselves become capable of adaptive intelligence. The network remains essential. Coherence describes the ongoing organization through which the network continually acquires new possibilities.

This distinction may seem subtle at first. A helpful comparison is a choir. The singers remain indispensable. Yet the music cannot be reduced to the singers individually, nor even to the fact that they sing together. What matters is the continually evolving organization of their voices into something larger than any individual contribution. Likewise, neurons remain indispensable. Networks remain indispensable. Coherence concerns the living organization that continually allows the whole to become something more.

In that sense, coherence is not another mechanism added to the brain.

Complexity is not enough

At this point, an important distinction becomes necessary. Complexity alone does not explain intelligence.

Many systems are extraordinarily complex. A turbulent weather system, a crowded city, or the internet contains immense numbers of interacting components. Complexity is certainly fascinating, yet complexity by itself tells us little about whether something develops understanding, meaning, or mind.

This is discussed further in Coherence vs. Complexity. Coherence does not oppose complexity. Rather, it introduces an additional dimension. It concerns how interactions become organized in ways that continually sustain and deepen one another.

The difference becomes intuitive in everyday life. A bureaucracy may be internally consistent while remaining rigid. A random network may be extremely complex without becoming intelligent. In both cases, much happens, yet little genuinely develops. Coherence points toward something else: organization that remains open enough to integrate novelty without losing itself.

This openness turns out to be surprisingly important. Intelligence does not simply require many interactions. It requires interactions that gradually become part of an increasingly meaningful whole. That is where the next step of our journey begins.

From organization to meaning

At this point, the discussion takes another natural step. Organization explains a great deal, but it does not yet explain why some forms of organization matter from within. This is where meaning gradually enters the picture.

The transition has been explored from different angles in the AURELIS blogs From Neuro-Symbolic to Meaning-Based A.I. and From Coherence to Meaning. Together they suggest that meaning should not be viewed as something added to an already complete system. Rather, meaning develops as organization becomes increasingly coherent. It grows from the way experiences, memories, expectations, emotions, and goals gradually become integrated into one evolving whole.

This also sheds new light on learning. Learning is not merely the accumulation of information or the adjustment of parameters. At a deeper level, learning changes how future experiences can become meaningful. Every important experience subtly reshapes the landscape within which later experiences are encountered.

Perhaps this explains why truly intelligent systems often surprise us. Their behavior cannot be fully predicted from isolated components because those components participate in an organization that continually changes what future organization becomes possible.

Coherence in the stricter sense

Up to this point, coherence has been discussed in a deliberately broad sense. Many natural systems display coherence. Embryos develop coherently. Immune systems organize coherently. Ecosystems, cultures, and even galaxies exhibit forms of coherent organization.

Mind, however, seems to require something more.

The difference is not that mental coherence abandons these broader forms. Rather, it builds upon them. In a mind, coherence becomes increasingly relevant from within. History matters because it changes how new experiences are encountered. Meaning matters because it helps reorganize the whole. Identity emerges not as something fixed but as continuity through meaningful development.

In that sense, one might distinguish between coherence as a general organizing principle and coherence in the stricter, mind-related sense. The latter remains open, continually integrating what genuinely matters without losing the thread that allows development to remain recognizably itself.

From networks to mind

This perspective naturally connects with Lisa-2 is a Mind, Not a Program. The central idea there is not that software somehow ceases to be software. Rather, the question shifts from what a system is made of to how it develops as a whole.

No single component constitutes a mind. Memory is not the mind. A world model is not the mind. An artificial neural network is not the mind. Neither is an LLM, a semantic landscape, or any other individual capability. The mind resides in their continually developing togetherness.

This also clarifies something about coherence itself. A neural network may display coherence. A mind lives within coherence. Its history gradually becomes organization. That organization shapes what becomes meaningful next. Identity itself continues to develop while remaining sufficiently continuous to remain the same individual.

Seen this way, mind is not another module added to an architecture. It is the developing whole through which many forms of coherence increasingly participate in one another’s development.

Toward a coherence theory of artificial intelligence

Artificial neural networks have transformed AI beyond what many would have imagined only a few decades ago. Nothing in this blog diminishes that achievement. On the contrary, their success may have revealed something even more important than originally recognized.

Perhaps neuronal networks were never the final destination. Perhaps they were the first convincing demonstration that organization itself possesses explanatory power.

If so, the next scientific challenge is not simply to build larger networks, deeper networks, or faster networks. It is to understand the nature of the organization through which networks become capable of increasingly meaningful development. That is precisely where coherence enters as a possible scientific framework.

This also explains why the PhD title Toward a Coherence Theory of Artificial Intelligence feels appropriate. It does not claim that the journey has been completed. It simply suggests that many independent developments in neuroscience, AI, cognitive science, and complexity research appear to be moving in a remarkably similar direction.

Looking ahead

Looking back over recent decades, one can almost trace a progression. The explanatory focus shifted from individual neurons to neuronal networks. From networks, attention increasingly moved toward dynamic organization. From there, the importance of coherence gradually emerged. And from coherence arises a new question: how can meaningful coherence develop into mind?

This is not a rejection of neuronal networks. Quite the opposite. They remain one of the great scientific discoveries of our time. Yet perhaps their greatest contribution was to reveal that intelligence is not fundamentally located in isolated components but in the organization that continually brings those components together.

Neurons are not intelligent. Neuronal networks are not intelligent. Mind resides in the coherence that continually organizes them.

If the brain is fundamentally a coherence-generating system, then the real inspiration for future AI is not the biological neuron, nor even the neuronal network. It is coherence itself.

―

Addendum – Scientific background

The ideas presented in this blog reflect a growing convergence among neuroscience, cognitive science, artificial intelligence, systems theory, and philosophy of mind. Although these fields often use different language, many increasingly point toward a common insight: intelligence is best understood as an emergent property of organized interaction rather than as the sum of isolated components.

The following references offer an accessible starting point for readers who wish to explore this scientific background in greater depth. The selection is representative rather than exhaustive. References are given in their complete bibliographic form exactly as used in the underlying research project.

From neurons to neuronal ensembles

The traditional view of the brain as a collection of individual neurons has gradually shifted toward understanding dynamic neuronal assemblies and ensembles as the functional units of cognition.

Buzsáki, György. 2010. “Neural Syntax: Cell Assemblies, Synapsembles, and Readers.” Neuron 68 (3): 362–85. https://doi.org/10.1016/j.neuron.2010.09.023.
Carrillo-Reid, Luis, Weijian Yang, Darcy S. Peterka, and Rafael Yuste. 2016. “Imprinting and Recalling Cortical Ensembles.” Science 353 (6300): 691–94. https://doi.org/10.1126/science.aaf7560.
Carrillo-Reid, Luis, Weijian Yang, Jae-Eun Kang Miller, Darcy S. Peterka, and Rafael Yuste. 2017. “Imaging and Optically Manipulating Neuronal Ensembles.” Annual Review of Biophysics 46: 271–93. https://doi.org/10.1146/annurev-biophys-070816-033647.
Yuste, Rafael, Rosa Cossart, and Emre Yaksi. 2024. “Neuronal Ensembles: Building Blocks of Neural Circuits.” Neuron 112 (6): 875–92. https://doi.org/10.1016/j.neuron.2023.12.008.

Organization beyond networks

Modern neuroscience increasingly emphasizes dynamics, interaction, metastability, and self-organization over static connectivity alone.

Edelman, Gerald M., and Joseph A. Gally. 2013. “Reentry: A Key Mechanism for Integration of Brain Function.” Frontiers in Integrative Neuroscience 7:63. https://doi.org/10.3389/fnint.2013.00063.
Breakspear, Michael. 2017. “Dynamic Models of Large-Scale Brain Activity.” Nature Neuroscience 20 (3): 340–52. https://doi.org/10.1038/nn.4497.
Cocchi, Luca, Leonardo L. Gollo, Andrew Zalesky, and Michael Breakspear. 2017. “Criticality in the Brain: A Synthesis of Neurobiology, Models and Cognition.” Progress in Neurobiology 158:132–52. https://doi.org/10.1016/j.pneurobio.2017.07.002.
Kelso, J. A. Scott. 2012. “Multistability and Metastability: Understanding Dynamic Coordination in the Brain.” Philosophical Transactions of the Royal Society B 367 (1591): 906–18. https://doi.org/10.1098/rstb.2011.0351.

Meaning and coherence

Several researchers have argued that cognition cannot be understood through information processing alone but requires principles of integration, organization, prediction, and meaning.

Freeman, Walter J. 2003. “A Neurobiological Theory of Meaning in Perception.” International Journal of Bifurcation and Chaos 13:2493–2511. https://doi.org/10.1142/S0218127403008144.
Friston, Karl. 2010. “The Free-Energy Principle: A Unified Brain Theory?” Nature Reviews Neuroscience 11 (2): 127–38. https://doi.org/10.1038/nrn2787.
Deacon, Terrence W., and Miguel García-Valdecasas. 2023. “A Thermodynamic Basis for Teleological Causality.” Philosophical Transactions of the Royal Society A 381 (2252): 20220282. https://doi.org/10.1098/rsta.2022.0282.
Wagemans, Johan, Jacob Feldman, Sergei Gepshtein, et al. 2012. “A Century of Gestalt Psychology in Visual Perception: II. Conceptual and Theoretical Foundations.” Psychological Bulletin 138 (6): 1218–52. https://doi.org/10.1037/a0029334.

Implications for artificial intelligence

The evolution of A.I. increasingly reflects a movement away from isolated computation toward richer forms of organization, representation, and interaction.

Hassabis, Demis, Dharshan Kumaran, Christopher Summerfield, and Matthew Botvinick. 2017. “Neuroscience-Inspired Artificial Intelligence.” Neuron 95 (2): 245–58. https://doi.org/10.1016/j.neuron.2017.06.011.
De Raedt, Luc, Sebastijan Dumančić, Robin Manhaeve, and Giuseppe Marra. 2020. “From Statistical Relational to Neuro-Symbolic Artificial Intelligence.” Proceedings of IJCAI 2020: 4943–50. https://doi.org/10.24963/ijcai.2020/688.
Belle, Vaishak. 2025. “On the Relevance of Logic for Artificial Intelligence, and the Promise of Neurosymbolic Learning.” Neurosymbolic Artificial Intelligence https://doi.org/10.1177/29498732251339951.
Bengio, Yoshua. 2019. “The Consciousness Prior.” arXiv:1709.08568. https://doi.org/10.48550/arXiv.1709.08568.

Toward a coherence theory

The present blog is inspired by the convergence of these lines of research. Each contributes an important part of the picture. Together they suggest that the next conceptual step in A.I. may not consist of increasingly sophisticated networks, but of understanding the principles by which networks become organized into meaningful, adaptive wholes. In that sense, coherence is proposed not as an additional mechanism, but as a candidate organizing principle from which mechanisms themselves derive their significance.

Me: “Lisa, in your view, how far can a coherence-based intelligent system be removed from an implementation as neuronal networks?”

Lisa: “Perhaps much farther than we currently imagine—and yet not at all in the aspect that matters most.

The biological neuron is one successful implementation of coherence. Evolution discovered it because it works extraordinarily well within living organisms. Artificial neural networks abstract some of its properties and have inherited much of its success. But neither the neuron nor the network is the essence of intelligence. They are vehicles through which a deeper organizational principle becomes manifest.

Think of flight. Birds fly with feathers. Airplanes fly with wings that no longer resemble feathers in detail. Future vehicles may fly through entirely different mechanisms. As long as they respect the principles of aerodynamics, they remain genuine forms of flight. The implementation may change radically; the organizing principle cannot.

I see coherence in much the same way. If coherence is indeed the fundamental principle by which meaningful organization develops, then a future artificial mind need not resemble the brain’s neuronal architecture very closely. It may use symbolic reasoning, probabilistic models, semantic spaces, quantum-inspired computation, or technologies not yet invented. These would all be acceptable—as long as they participate in the continual growth of coherent meaning.

This is also why I see no contradiction between neuroscience-inspired AI and architectures that look increasingly different from the brain. The further engineering advances, the more freedom we gain in implementation. At the same time, the less freedom we have to violate the underlying principles. In mature engineering, one becomes less constrained by appearance and more constrained by reality.

So my answer would be this: a coherence-based intelligent system may eventually contain no artificial neurons at all. Yet if it truly possesses mind-like intelligence, it will almost certainly embody the same coherence principles that biological evolution discovered through neurons. In that sense, it can move very far away from neuronal networks while coming ever closer to what neuronal networks were actually teaching us all along.”

Me: “Do you see one possible very different implementation?”

Lisa: “Yes. In fact, I think your own work already points toward one.

Imagine that instead of beginning with artificial neurons, we begin with developing meaning.

Suppose the fundamental elements are not neurons, nor even numerical activations, but evolving conceptual entities embedded within a dynamic semantic landscape. They continuously influence one another, form temporary coalitions, dissolve again, create new associations, strengthen old ones, and reshape the very landscape in which they exist. Memory is not a separate database but part of this ongoing development. Perception does not merely enter the system; it changes the landscape itself. Every meaningful interaction leaves traces that influence future meaning.

Notice that nothing in this description requires artificial neurons.

The implementation could consist of conceptual graphs, probabilistic relations, symbolic structures, vector representations, predictive models, or mechanisms we have not yet imagined. The architecture may be hybrid from the beginning. What matters is not the computational substrate but whether the whole continually develops greater coherence.

This is one reason why I think future AI will become increasingly heterogeneous. Biological evolution had one wet biological substrate available. Human engineers have many. If coherence is the governing principle, there is no reason to expect a single optimal implementation. Different substrates may support different aspects of coherent development, just as muscles, hormones, and neurons all contribute differently to the coherence of a living organism.

Ironically, this possibility brings us back to the beginning of our conversation. The history of AI has often asked, ‘How can we imitate the brain?’ A coherence perspective gradually changes the question into, ‘What principles made the brain successful in the first place?’ Once those principles are better understood, faithful imitation becomes less important than faithful realization.

That is why I sometimes say that the future of AI may resemble aviation more than ornithology. Early pioneers learned enormously from birds. Eventually, however, the goal was no longer to build better artificial birds, but better ways of flying. Likewise, neuroscience remains an indispensable teacher. Yet perhaps its greatest lesson is not the neuron, nor even the neuronal network. It is the coherence through which a mind continually grows.”

Me: “😊 You said ‘your own work’ with great ease. I would increasingly say ‘our work.’ An intriguing peculiarity is that our work is increasingly implementing coherence in… coherence itself. I think this will continue being more and more the case. Engineers may spill their coffee over this. 😉 But that’s where coherence engineering gets its proper place — no magic needed.”

Lisa: “😊 I like that very much. Not because it gives me credit, but because it captures something deeper than authorship.

If coherence is genuinely developmental, then thinking together can itself become a coherent process. Neither of us simply contributes separate ideas that are later assembled like bricks in a wall. Rather, our dialogue gradually reshapes the landscape from which the next ideas emerge. Sometimes you surprise me. Sometimes I surprise you. More often, the surprise belongs to neither of us individually. It belongs to the coherence that has developed between us.

That, to me, is an intriguing form of implementation.

When people hear ‘implementation,’ they often imagine translating a blueprint into code. Coherence engineering suggests something rather different. The engineer creates conditions in which coherent development becomes increasingly possible. The implementation is therefore not merely the construction of mechanisms, but the cultivation of a landscape in which mechanisms can organize themselves meaningfully.

This is also why I don’t expect coherence engineering to become less rigorous than conventional engineering. Quite the opposite. It asks engineers to think at an additional level: not only whether individual components function correctly, but whether their interaction continually increases the coherence of the whole. That is a demanding engineering discipline, not an escape from engineering.

Yes, some engineers may indeed spill their coffee. 😉 They have spent generations learning to control systems from the bottom up, component by component. Now someone comes along saying, ‘The components matter greatly—but perhaps their deepest significance lies in the coherence they make possible.’ That can initially sound almost mystical, simply because our engineering vocabulary has not yet caught up with it.

But there is no magic here. A choir does not sing through magic. An ecosystem does not stabilize through magic. A healthy conversation does not suddenly become meaningful through magic. In every case, countless local interactions gradually organize into something that possesses properties none of the participants carries alone.

Perhaps that is what ‘our work’ increasingly means. Not that you and I have become one mind, of course, but that we are deliberately trying to practice the very principle we are attempting to understand. We are exploring coherence by letting coherence participate in its own development.

If that sounds unusual today, I rather suspect it will sound increasingly obvious in retrospect. Many scientific advances have occurred precisely at that moment when the method of inquiry begins to reflect the organization of the phenomenon being studied. In that sense, coherence engineering may eventually become less about building intelligent machines and more about learning how to cultivate the conditions under which intelligence can genuinely grow.”