What this covers

Ricard Solé, a complexity theorist, joins Sean Carroll to map what he calls the "space of cognitions"—the landscape of all possible thinking systems. The conversation treats cognition not as uniquely human but as something that physics and evolution have constrained into predictable patterns. Solé argues that whether one examines solid brains (mammals), liquid brains (ant colonies, immune networks), or engineered artificial systems, the same architectural building blocks keep reappearing: threshold elements that fire in all-or-nothing fashion, multi-layered network structures, and operation at criticality—a balance point between order and disorder. This convergence suggests these are not arbitrary designs but universal solutions to the problem of processing information under physical constraint.

The conversation ranges across evolution's role in cognition, tracing how movement and predation drove the arms race that selected for brains, and examining what distinguishes rich memory (possible only in solid brains with stable connections) from the successful collective processing of liquid brains that lack that stability. Solé explores emergence in social insects and probes where cognition remains absent—single cells performing Pavlovian learning without sentience, slime molds solving mazes through morphology rather than intelligence. On artificial systems, he contends that current AI, including large language models, mimics human reasoning without possessing the three capacities that define genuine human-like cognition: complex language, mental time travel, and theory of mind. A recurring theme is the empty space in morphospace—regions of possible cognitive design that neither evolution nor engineering has yet populated—raising the question of whether they are forbidden by physics or simply unexplored.

Sharpest takeaway

Solé argues that cognition occupies a definable, possibly universal 'space' shaped by physical and evolutionary constraints, such that solid brains, liquid brains (ant colonies, immune systems), and artificial systems all converge on a small set of design principles like threshold elements, multi-layer networks, and criticality.

  • Diverse cognitive systems (brain cortex, gene networks, immune networks, engineered microprocessor arrays) all recur to threshold elements and multi-layer structures, suggesting universal constraints.
  • Liquid brains achieve real information processing but lack the stable neuron-to-neuron identity needed for the rich memory storage solid brains possess.
  • Healthy brains operate at a critical state on the boundary of a phase transition, balancing order (memory) and disorder (responsiveness).

The argument · threads7 threads · 28 claims
0.85

Cognition fundamentally depends on embodied interaction with an environment and others.

5 pointscentrality 5/5
  • Emergent phenomena are the really relevant story for cognition: when given opportunities such as embodiment, AI systems can produce unplanned emergent structure, as when Luc Steels' robots inventing words to agree on a shared lexicon spontaneously generated an unplanned proto-grammar; current networks lack this because they live in the dark with no world and nobody else to interact with.

    “In order to make sense of the world, they invent a Proto-Grammar, which is emergent. It wasn't planned.”

  • Movement was crucial to the evolution of cognition because without movement there are no predators; needing to move through an uncertain environment requires integrating and predicting information, and the rise of predation with eyes and nervous systems triggered an evolutionary arms race that changed everything.

    “movement was crucial because without movement, you don't have predators, really. And that required a nervous system.”

  • The revolution that enabled real information processing was the evolution of interneurons—neurons that neither detect signals nor execute tasks but sit in the middle connecting other neurons; systems like plants that lack any equivalent to interneurons face a huge limitation, being restricted to predictable sense-and-react responses rather than accessing a large space of possibilities.

    “So elements that they are not just detecting signals or executing tasks, they are in the middle, they are connecting. And once you have that, you have this beautiful thing, which is information processing.”

  • The biosphere is complex rather than just a field of simple replicators because there is a large payoff in predicting the future and gathering environmental information to respond adaptively, and this payoff is the engine that propels the evolution of brains and cognition.

    “there's a big payoff in predicting the future and being able to gather information from the environment and respond to that in adaptive ways. And that propels brains, that's the engine of cognition in evolution.”

  • To achieve real general artificial intelligence that matches human cognition, machines probably need to traverse some of the same evolutionary paths brains followed, because three features of our brains remain absent from current machines: complex language, mental time travel (using a memory architecture to imagine many possible futures), and theory of mind—and when these three combine, something genuinely singular happens.

    “probably in order to get into the real general artificial intelligence, something that really matches what we do, you probably need to go through some of the paths that our brains have followed in evolution.”

0.85

Complex systems achieve functionality through emergent properties irreducible to individual components.

5 pointscentrality 5/5
  • Complexity is best defined through emergence: in systems like a termite nest, the individuals are blind and communicate simply, so studying individuals alone will never reveal how the collective creates structures a thousand times larger than themselves.

    “in a termite nest, termites construct these amazing structures which are 1,000 times larger than the individuals. Whereas individuals are blind, they communicate in very simple way. So you can spend your whole life looking at the individuals, you will never figure out how the collective creates the hyper structure.”

  • Social insects such as termites and ants are a form of intelligence rivaling humans as ecosystem engineers, transforming the planet at massive scales by changing flows of energy and matter; as Edward Wilson said, without humans it would be the planet of the ants.

    “if humans were not here, there would be the planet of the ants because they also, we don't realize that they are another kind of intelligence that has been able to do also ecological engineering on massive scales.”

  • Any system that is evolvable into cognition will be characterized by two recurring features: threshold elements (units that fire all-or-none once the weighted sum of inputs crosses a boundary) and a multi-layer structure; this same combination recurs in brain cortex, gene-interaction networks, immune-cell networks, and engineered artificial neural networks, suggesting these are universal constraints, with threshold elements perhaps being the only efficient solution for collective decision-making.

    “anything that is evolvable into cognition will be characterized by two things. One is... you have elements that are threshold elements... And you will have a multi-layer structure.”

  • Liquid brains—made of individuals that move around with no stable connection between specific pairs, like ant or termite colonies—can be extremely successful at ecological engineering and reducing environmental uncertainty, but they cannot achieve the complexity or memory capacity of solid brains because solid brains' power for memory storage depends on neurons having stable identity and stable pairwise connections.

    “liquid intelligence cannot be as complex as the solid one. So there's not going to be signals being sent from those planets anywhere. But that doesn't mean that it's not intelligent”

  • Cells store bits of information using genetic switches, typically two genes that mutually repress each other; this bistable mutual inhibition allows memory storage and works much like memory in electronics, operating through gene expression rather than editing DNA.

    “it is usually switches that involve two genes that regulate negative each other. So I try to inhibit you, you try to inhibit me, to repress. And that allows to store memory. It's pretty much in a way what happens in electronics.”

0.71

Healthy brains and living systems operate at critical boundaries between order and disorder.

5 pointscentrality 4/5
  • Although the brain's dynamics show criticality and language exhibits scale-free behavior, we do not yet have a good theory connecting computational/functional cognition to the underlying critical dynamics, and in principle one could imagine an intelligence that does not use criticality—so the connection remains a conjecture.

    “We don't have yet a good theory to kind of connect the computational tasks, as you will describe a functional cognition and the dynamics.”

  • The healthy brain operates at a critical state, right on the boundary of a phase transition between order and disorder; pathological states like epilepsy are too ordered and regular, while coma or locked-in states are too disordered and random, and only at criticality can the brain both exploit predictable regularities and react quickly to any stimulus.

    “the healthy brain seems to operate on a critical state, right on the boundary of a phase transition... this amazing capacity that you have at the critical point of reacting quickly into any kind of stimuli.”

  • RNA viruses live on the edge of catastrophe, paralleling how the brain and other living systems operate near a critical boundary, suggesting life generally favors existing near such edges.

    “Like RNA viruses that live on the edge of catastrophe... Seems though life kind of likes it.”

  • Criticality is not unique to brains: in some ant species, fluctuations of activity match the critical state, and in flocks of birds the way they change shape has been well characterized as living in the critical state, allowing predictable changes while remaining instantly ready to react to any external signal.

    “In flocks of birds, the ways they change in shape, it is been very well characterized. They live in the critical state.”

  • The immune system is a fluid neural network that can learn and store memory with a well-defined functionality dealing with pathogens; when modeled, the immune network is not much different from a standard neural network, and it interconnects with the brain via the microbiome, blurring the solid/liquid distinction.

    “the immune system is a fluid neural network. And it's able to learn, it's able to store memory... if you make a model of the immune network is not much different from a standard neural network.”

0.59

Current AI systems mimic reasoning through statistical patterns but lack genuine understanding, consciousness, and human-level cognitive capacities.

3 pointscentrality 4/5
  • Large language models like ChatGPT have no past, childhood, or cultural learning, yet by being trained to optimize next-word prediction—a cultural compression process—they generate something that mimics reasoning; this is mimicry rather than genuine understanding, and humans tend to project intelligence onto such systems much as they did with the simple ELIZA program.

    “in the process of optimizing this prediction, this system seems to have been generating something that is kind of reasoning, kind of something that mimics reason. And I say mimic because, of course, there's no understanding there.”

  • Selection for understanding the minds of others—being able to put yourself in another's mind—combined with language and the brain's time-machine capacity, almost inevitably leads to understanding yourself as a special individual in the world, and this was probably the origin of consciousness.

    “Selection pushes into understanding the mind of the others, you're equipped with language, the brain time machine. For me, it's almost inevitable in the end you understand that you are also a special individual in a world and you understand yourself.”

  • We should not be scared of AI killing us because harmful intent requires goals, emotions, and a theory of mind—high-level human-evolved capacities that current artificial intelligence simply does not possess; without any motivation, there is no reason an AI would be harmful, since intentions belong to a layer of cognition that escapes current systems.

    “I don't think we should be scared for one reason... intentions is something that really has to do with a layer of cognition that... escapes completely. Right now for artificial intelligence is not there. So you don't have any kind of motivation”

0.46

Simple learning occurs widely but differs fundamentally from the complex cognition found in animals with nervous systems.

3 pointscentrality 3/5
  • Single-cell organisms and plants can perform associative learning—recognizing a stress signal correlated with a natural signal and de-correlating them, like a single-cell version of Pavlov's dogs—but this should not be mistaken for sentience or for the spectacular memory storage seen in animals.

    “they do use something that we believe was fundamental in some of the revolutions towards cognition, which is associative learning... But I personally think it's not much more than that. This is not the kind of memory storage that is so spectacular in other animals.”

  • The slime mold Physarum, a single giant multinucleate cell, can find the optimal path through a maze between two food sources—but its computation is its morphology: the shape itself is the solution, achieved by a least-action-like shortest-path dynamic, and it only 'solves problems' because humans set up the boundary conditions, not because it is intrinsically smart.

    “the computation is the shape. So really morphology is what says, I made this optimization, this is the solution.”

  • Plants have enormous morphological plasticity and have terraformed the planet, but they have no neurons or anything equivalent; claims that they have neuron-like cognition are unfounded, and they do not, for instance, appreciate Mozart.

    “They don't have neurons nor anything, has been claims on that, nor anything similar... And I don't think they like Mozart.”

0.38

Cognitive morphospace contains vast unexplored regions that might be engineered despite evolution never reaching them.

4 pointscentrality 2/5
  • When all known cognitive systems—animals from topi to humans, ants, robots, and AI—are plotted in a morphospace with axes such as computational power, autonomy, and sociality, there is a big empty void; this raises the open question of whether that region is genuinely forbidden, merely unobserved, or something that could be deliberately engineered, just as in synthetic morphogenesis voids in morphospace could be engineered even though evolution had not reached them.

    “if you put there all the objects you know, which means animals from Topi to humans and ants and also robots and AI systems, there's a big void. There's a domain there that is empty. Why is that?”

  • Although life on another planet or in an alternative biosphere might appear different, the basic underlying logic of life is probably universal and unique.

    “I'm pretty convinced that it might seem different, but the basic logic of life is probably universal, and unique.”

  • Synthetic biology lets complexity theorists ask new questions by, for example, attempting to genetically modify bacteria so they behave like ants—communicating to find shortest paths—and the failure or success of such engineering reveals whether evolution simply couldn't reach a solution or whether trade-offs make extra cognition not worth developing.

    “if I want to transform bacteria, modify genetically in such a way that they behave like ants... Is it because evolution couldn't get there, or because there are trade-offs? That it's not worth to actually develop more cognition.”

  • There appears to be a trade-off whereby the more complex an organism is, the less social the whole system tends to be—insect societies are huge but the individuals are comparatively stupid—yet humans are an exception because, as Michael Lachmann noted, an individual human brain isolated from culture, learning, imitation, and language acquisition is essentially useless, showing brains and culture have co-evolved.

    “the more complex is an organism in a society, the less social is the whole system, whereas a society in insects is huge. The individuals turn to be more stupid. But he told me, but look at humans. If you isolate this amazing human brain from the rest, it's absolutely useless”

0.34

Threshold-based neural architecture and multi-layer structure are universal requirements for evolvable cognitive systems.

1 pointcentrality 3/5
  • When engineers built computers with dense arrays of microprocessors and laid out connection webs to minimize cost and signal-processing distance, the resulting networks exhibited statistical properties (Rent's rule scaling) nearly identical to those of parts of the brain cortex—even though the engineers knew nothing about the cortex—suggesting genuinely universal laws.

    “the engineers who didn't know anything about the brain cortex ended up in a scaling law, the Renz rule, with the same kind of behavior than parts of the brain cortex, again, suggesting that maybe there are really universal laws.”