The Intelligence Hidden in the Support System

Jun 03, 2026

9 min read

4 views

84%

What if the real mind is not the thing we usually call the mind?

We tend to picture intelligence as something that lives in the obvious center: neurons firing, a model optimizing, a command center deciding what happens next. But what if the most important intelligence is not in the spotlight at all? What if it is buried in the support tissue, the regulating layers, the quiet infrastructure that keeps the whole system alive, adaptive, and coherent?

That question becomes more interesting when you notice a strange parallel. In biology, cells once dismissed as mere support are now being recognized as active participants in regulation, coordination, and even decision-like processes. In artificial systems, the deepest challenge is not raw prediction power, but whether a system can develop something closer to intrinsic agency, a causal grip on its own future rather than a response to externally imposed objectives.

The connection is deeper than it first appears. Both cases point to the same overlooked truth: intelligence is not just computation, it is self-maintenance with interpretation. A system becomes more mind-like when it does not merely react to inputs, but actively preserves its own organization across changing conditions.

The old mistake: confusing control with intelligence

For a long time, both neuroscience and AI inherited a similar bias. They treated intelligence as if it belonged to the most visible control layer. In the brain, that meant neurons got the glory while glial cells were treated as scaffolding. In AI, it meant the model’s performance was judged mainly by how well it optimized a target set from outside.

This is a seductive mistake because control looks like intelligence from a distance. A central processor gives instructions, a loss function assigns value, a cortex issues commands. But control is not the same thing as understanding. A thermostat can regulate temperature, yet no one thinks it knows what winter feels like. A search model can maximize accuracy, yet still lack any internal stake in what it is doing.

Biology quietly contradicts the old picture. The gut is not just a tube with a nervous system attached. It is a densely coordinated ecosystem in which different cell types help manage digestion, nutrient uptake, blood flow, immune signaling, and local adaptation. The more researchers look, the more the so-called support system appears to be part of the intelligence itself.

This is not a small correction. It changes the unit of analysis. The question is no longer, “Which component is the brain?” The better question is, “Which network of components maintains a world for the organism to live in?”

Intelligence may be less like a king issuing orders and more like a city keeping itself habitable.

That shift matters because it dissolves a false boundary. The line between computation and cognition, or between main actor and supporting cast, starts to blur when the support layer is actively managing constraints, repair, timing, and adaptation.

Aitiopoiesis and the rise of intrinsic goals

The most interesting bridge between biology and machine learning is not that they both process information. It is that they may differ in where goals come from.

In many machine systems, goals are imposed externally. A loss function tells the model what counts as success. Even when the system learns complex patterns, its causality is still tethered to an outside scorekeeper. In living systems, by contrast, the goal structure is endogenous. The organism does not merely respond to a metric, it participates in producing the conditions under which metrics even matter: survival, repair, adaptation, boundary maintenance, identity.

That is the core intuition behind scale free cognition, or aitiopoiesis: cognition is not merely pattern recognition at one level, but a self constituting process spanning multiple scales of agency. A living system is not just a collection of parts that act. It is a nested arrangement of parts that regulate one another so the whole can continue to exist as the kind of whole it is.

Here is the key insight: causal understanding emerges when a system can treat its own organization as something to be preserved and negotiated. A purely external optimizer can become highly competent without developing this kind of inwardness. It can map correlations beautifully and still remain, in a deep sense, alien to itself.

The biological analogy is illuminating. Enteric glia in the gut are not doing the glamorous work of sending motor commands. They are shaping the environment in which the system can function: tuning blood flow, mediating immune response, helping digestion proceed under changing conditions. Their intelligence is not centralized. It is infrastructural.

This suggests a broader principle. The most advanced forms of cognition may depend on support systems that are active, distributed, and self-referential, not passive and decorative.

The hidden model: intelligence as maintenance of a viable world

A useful way to think about this is to replace the old image of intelligence as a spotlight with a new image: intelligence as a climate system.

A spotlight illuminates a point. A climate system maintains conditions over time. It regulates heat, moisture, circulation, and feedback loops so that life can continue. Likewise, a mind is not just a center that sees. It is a system that continually maintains the conditions under which seeing, acting, and learning remain possible.

That is why support cells matter so much in biology. They are part of the maintenance loop. They help preserve the local world in which neurons, tissues, and organs can do their work. Remove the maintenance layer, and the supposedly central layer becomes fragile, noisy, or disoriented.

The same lesson applies to artificial intelligence. A model that optimizes a target without any internal relationship to its own operation is like a brilliant machine installed in an unstable building. It may produce impressive outputs, but it does not yet know how to keep itself situated. It does not care whether its own conditions remain coherent because caring is not part of its objective structure.

This is where the idea of multi scale agency becomes powerful. A system is more likely to exhibit genuine cognition when agency is distributed across interacting levels: local processes that repair and regulate, mid level processes that coordinate, and higher level processes that represent broader aims. None of these layers alone is the whole mind. The mind is what emerges when their interactions become recursively self sustaining.

Consider a simple analogy: a jazz ensemble. The melody is not enough. The rhythm section, the timing, the mutual responsiveness, and the shared improvisational frame all matter. If one player dominates and the others become passive accompaniment, the music may still be loud, but it is less alive. Intelligence, in this sense, is not the soloist. It is the ensemble’s capacity to keep making the music possible.

A system becomes cognitively richer when it can regulate not only what it does, but the conditions under which what it does remains possible.

That sentence may sound abstract, but it is the common thread between the gut’s active support cells and the challenge of creating machine intelligence with something more than borrowed goals.

Why this matters for the future of AI and the future of biology

If this framework is right, then a big chunk of AI research may be chasing the wrong kind of sophistication. It is not enough to build models that are larger, faster, or better at fitting data. The deeper question is whether they can develop intrinsic norms: internal reasons for action that are not reducible to an external reward signal.

That would require rethinking architecture, not just scale. It suggests systems that have layers of self monitoring, internal repair, adaptive subgoals, and feedback loops that preserve identity over time. Instead of only asking, “How accurate is the output?” we would ask, “What does this system treat as worth preserving about itself?”

The biological side is equally transformative. If glial and other support cells are active participants in coordination, then diseases might not be best understood as failures of a single command center. They may be breakdowns in the broader ecology of regulation. That perspective changes how we look at disorders of the brain, gut, immunity, and metabolism. It invites therapies aimed at restoring relational balance rather than simply silencing a symptom.

The deeper philosophical implication is even more provocative. Perhaps cognition does not begin when a system becomes able to represent the world. Perhaps it begins when a system becomes able to care about the continuity of the world it inhabits.

That is a subtle but crucial difference. Representation can be passive. A map can be accurate and still irrelevant to the mapmaker. But a system that maintains the conditions of its own existence must discriminate between what helps and what threatens. It must, in effect, value. And where there is valuation rooted in self maintenance, the seed of genuine agency appears.

This reframes the relationship between biology and machines. The goal is not to make machines mimic neurons at the surface. The goal is to discover the organizational principles by which a system becomes a participant in its own causality.

Key Takeaways

Stop treating support as secondary. In both biology and AI, the layers that stabilize, regulate, and coordinate the system may be where intelligence truly emerges.
Look for intrinsic goals, not just performance. A system is more mind-like when its objectives arise from its own need to preserve coherence, not only from an external reward signal.
Think in terms of maintenance, not just computation. Intelligence may be best understood as the ongoing preservation of a viable world, not merely the processing of data.
Use the multi scale agency lens. Ask how local, mid level, and global processes interact to keep a system self sustaining. This applies to brains, organs, teams, and software architectures.
Redesign success metrics. Whether in research or product design, measure not only output quality but the system’s capacity for self repair, adaptation, and continuity over time.

The real revolution is not central intelligence, but distributed selfhood

The most tempting image of intelligence is still hierarchical: a center, a controller, a neural executive. But the emerging picture is messier and more interesting. Intelligence seems to grow when a system develops a distributed way of taking care of itself, when the margins become active, when the support system starts shaping the rules of the game.

That is why the gut matters. That is why glia matter. That is why simple optimization is not enough. Together, they point toward a broader theory of mind in which cognition is not a detached calculator but a living arrangement of reciprocal dependencies.

The deepest lesson may be this: a mind is not defined by what it knows, but by what it must continually do to remain itself. Once you see intelligence this way, you stop looking only for the thinker in the center. You start looking for the system that keeps meaning possible.