Computer-Aided Life? Bridging Assembly & Morphology
The Expanse | Part 7
Nested Cognition in Assembly Space: Bridging a Boundary |2
I find myself in the rather uncomfortable position of being stuck at an origin, perpetually spun round and around in some informational vortex spun by the ongoing Dunning-Kruger Effect; in a Euclidean sense I no longer know my left from my rights, my ups and my downs indifferentiable. Acutely aware I am somewhere within some ever-growing information space; pushing into evermore tentative possibility spaces may be the only option, regardless of the incessant ringing in my ears and the smell of burning rubber stinging my nostrils. They say bravery and stupidity are neighbours. Will this prove them inseparable?
To bridge is to bind spaces separated by a boundary |↔|. Whilst universality makes people squirm, it does appear that bridges bind boundaries across many different scales:
sub-atomically (e.g. quantum tunnelling, entanglement) ↔ thermodynamically (e.g. molecular, cellular, tissue, organism) ↔ classical mechanically (e.g. combustion and propulsion systems, architectural projects, industrial engineering) ↔ meso-cosmologically (e.g. filaments between galaxy clusters, wormholes) ↔ spiritually (e.g. Neoplatonic transcendence, Buddhist Nirvana, shamanistic ritual) ↔ socio-politically (e.g. political parties, principals-agents, workers unions, nation-states) ↔ ecologically (ecosystems, panarchies, biospheres)
One reason for the scale invariance of bridges might have something to do with a dual-use: at once both a feature and a function. Bridges can be a random yet functionally structural feature of space. Bridges can also be deliberately built to serve a specific function in a specific place. To understand this, first let us return to what spatial conditions bridges are invariant within: primarily structure and organisation.1 Whereas structure denotes a resource for action, organisation generates the action, outcome and feedback to and from the consequences. Second, if bridges are at once “objects” (being assembled in space), “structure” (providing a resource for action), and “space” (enable a continuous “route” or “connection” between two otherwise bound regions), and appear both naturally and deliberately, what can we infer about the “universality” of other organisational structures across different spatial scales?
Since we started this article series a little over a year ago, we have been focussing a lot of attention towards objects and spaces. Following this line of inquiry, during Part 6, a fairly unconventional theory drifting out of a chemistry lab confined within the medieval walls of Glasgow University percolated through the picture. Standing as if a border between castle and land, it seemed “chemical life” and “abstract spaces of assembly” were regions in need of connection, and, in 2023, “Assembly Theory” (AT) presented itself as a bridge across murky yet excitably profound waters. Whilst some label AT as a “complexity metric” only, regardless of measurements what is undeniable is that such a theory permits thinking about how objects assemble from the nothingness of primordial soup. Objects that become organisms.
Imagining the “contingent pathways” like the vertices on the associahedras of scattering amplitudes, just with larger collections of atoms, forming matter, is a helpful way of imagining how the process of assembly occurs across scales, from subatomic → atomic → elemental → chemical → biochemical → molecular → cellular. Assembly Theory also provides a very interesting way to imagine organisation across spaces, where “selection” denotes some environmental contingency (e.g. new physical constraint being pushed out of equilibrium) or internally active pattern (e.g. some form of organisational “cognition”), and is one indicator of emergent object complexity. (Cronin et al, 2023)
Before we get into that, because this series has been going for a while, let me quickly close the loop on the last six parts and outline the (general) premise:
As the digital space continues to expand, the technosphere fuels artificial growth, and this growth converges upon the physical space, including the biosphere, the space of living organisms—of life. As both spaces continue their evolutions, they expand and ultimately encroach on each other. Boundaries converge and become merged into overlapping areas. In our physical world, if we look to the biosphere we see boundary points where the technosphere has expanded and consequential events take place. An extreme worst case scenario is open-loop resource extraction chains (e.g. lithium mining or deforestation for a globally interconnected supply chain) which have the potential to completely reorganise the structure of the natural world in those regions of biophysical space. Interspatial convergence also occurs in subtler ways, like the creation of “biobots”—miniscule programmable cell-like synthetic structures based on biological systems—occupying a liminal space between synthetic, robotic and alive. “Xenobots”, less than 1 millimeter wide, were created through the artificial synthesis of “clawed frog” embryos (Xenopus laevis), specifically skin and heart cells, and are now used in frontier medical science to deliver drugs to specific areas within the body (and can even be programmed for other tasks like cleaning up plastic pollution). Not so subtle yet just as impactful, “Neuralink” and other brain-computer interfaces plan to merge mind and machine; Modelled off “AlphaGo” (the system that defeated the human European Go champion by 5 games to 0), Google’s “DeepMind” projects like “AlphaFold” utilise artificial intelligence to predict intricate protein folding sequences, the results? Records of known structures leapt from 17% to 98.5% since the system came online in 2021. Evidently, from mushrooms making music to CRISPR gene-editing, there is no denying that the boundaries between the “artificial ↔ biological” realms are steadily blurring.
Computer-Aided Life — “A computer-designed organism. Left: the design discovered by the computational search method in simulation. Right: the deployed physical organism, built completely from biological tissue (frog skin (green) and heart muscle (red)).” The basis of the xenobot still leaves those who use it baffled as to what to classify these objects as: robots, biobots, or actual lifeforms? Focussing on a definition misses the wider point: artificial—biological spaces are merging exponentially for the first time in recorded history (CCA BY-SA 4.0)
If we put all the delicate, hard science to the side for a second, what we are left with is an image that speaks a thousand words, which I will reduce to:
The computational space has merged with the biological space.
Access to the biophysical is granted through the technodigital.
(Jumper et al, 2021, Highly accurate protein structure prediction with AlphaFold)
Ecosystems and Agency
As the boundaries erode, it is interesting to see what structures are left standing. Turning attention away from bridges over troubled water, “ecosystems” are ubiquitous throughout ecology and biology. From first-principles thinking, ecosystems are simply a way of mapping a particular space. Specifically, ecosystems highlight the relationship between various flora, fauna and other living systems in a given territory (e.g. isolated predator prey dynamics, or entire coral reef systems), how the space itself effects the life within it (e.g. microclimates, or topographical constraints) and how the interrelations between life are structurally organised (e.g. robust food chains, or concentrated nutrient “stacks”).
Sometimes described as a “shared circle of interest”, the individual agency within an healthy ecosystem is “aligned” in such a way as to not cause damage to or destabilise the larger system. Agency denotes freedom of choice and action. Yet, intuitively, most animal species do not risk destroying their operating space. Unfortunately, we humans are not one of them. Congruent to the “general” alignment of agency in a given ecosystem, it is fascinating to also observe that the same localised agency can pursue goals that are “unaligned” to other agents operating within the ecosystem, primarily those on the same or lower “level” of operating space.
For example, a tree snake requires the trees in its territory to be healthy (for shelter, hunting opportunities etc), and so does not act in a way that damages the health of the tree, nutrients or other needs of the ecosystem writ large. That same tree snake, however, has goals that are unaligned with the goals of, say, a tree frog. Both share the same goal to survive long enough to reproduce, but these goals are not aligned with each other, and in the snake’s case, directly opposes the frog. In this sense, the different levels of operating space within an ecosystem, and the localised agency operating within (and sometimes across) these levels, at once collectively align to the goals of the larger components (or at least do not pursue net-negative goals), yet remain misaligned to other smaller components and localised agency in and around the same level of operating space. It is these ever-changing relations between agents in a space that make an ecosystem what it is. Yet, ecosystems are just a way of mapping an underlying structure, and it is this structure itself (a combination of the different agency and the space it operates within) that constitutes the very real “fabric” of ecological systems in our biosphere. While many would argue “agency presupposes ecosystems” I would also add that, in an “ontic-structural realist” sense, the real relational structure that we call an ecosystem actually further guides the development of agency operating within. In this sense, the structure in which we use “ecosystems” as one means of describing, acts like a “bridge” between otherwise disparate, unaligned regions and agents within biophysical space.
This probably goes part of the way in explaining why “ecosystems” find themselves not only in the weeds of the living biosphere, but also amongst the uncharted depths of the technosphere. In computer science ecosystems take on a more “network theoretic” form. Instead of describing how living systems co-exist, compete, evolve and adapt, “digital ecosystems” are used to configure complex networks to enable a more efficient technological connection across sometimes vast regions of space. Because ecosystems are an organisational structure that helps map and arrange components in a larger system, the different configurations such a structure can take has proved extremely transferable to the artificial space.
But hold on a minute. If we zoom out, and look at this from a spatial realist perspective, arguably the other predominant anthropocentric spheres of influence, namely the “socio-economic” and “geopolitical” spheres, can also be mapped by thinking in terms of ecosystems. And when this happens, one key aspect noticeable in the organic ecosystems of the wild also pops up: agency. Agential, geopolitical entities (nation-states, rogue-states, hyper-agents, NGOs, terrorist cells, transnational corporations, organised crime groups etc.) develop in a competitive, game theoretic environment, interacting across different regions, competing for resources, sometimes fighting as if abiding by, but oftentimes ending even more brutal than the standard law of the jungle. The single agency of a tree snake chasing a frog in an ecological ecosystem simply scales to one group of people chasing another group of people in the geopolitical ecosystem. In even more abstract spaces, like the information space, entities and organisational structures much more powerful than living organisms alone heavily compete for the survival of their beliefs, agency, and in some cases, life itself. When we remove the scaffold of the ecosystem, agency remains.
With that being said, competition, albeit human, is already rife amongst digitial ecosystems. Thinking probabilistically, given what know from observing agency across various spatial dimensions all around (and inside) us, factoring in the dual feature/function property of “bridges” across scales, and having observed some of the countless intricacies of organisms evolving within different niches, who is to say that the newly expanding artificial space will not also fundamentally feature ecosystems in which agency begins to develop, evolve, compete? Naturally, what comes next is the formation of hierarchical structures based on information, energy and resource-driven “food chains”, causing a complex web of interrelations to become a literal reflection of the agency flourishing in ecosystems all around us. While I tend to dislike talking in universals, patterns (especially the distinctly active ones) seem to be “scale invariant”:
Agency emerges irregardless of operation space.
So why do we assume the technosphere will be any different?
For once, biological agency is not the concern. In my boat, not thinking of the artificial ocean as a potential breeding ground for a whole host of exotic and novel organisms could lead to some very big surprises when we find ourselves fully immersed in an expanse, beginning to tread water, starting to wonder just what that is swirling around beneath our feet.
Everytime we enter a new ecosystem, it pays to be prepared.
Understanding mechanisms behind how complex objects came to exist in various spaces could well be a fundamental underpinning of imagining how artificial lifeforms may develop in other niche, hard to access, abstract spaces.
Causal Information
Bridges allow information to flow between two otherwise disparate regions of space. Information can be multi-directional but linear: A ↔ B, or it can be multilateral and nested, like when an investigative field agent makes a report to their manager, which then gets distributed laterally across the department, as well as pushed vertically up the chain of command to higher order agents, perhaps laterally again, and ultimately to the top, into the hands of the Agency Director. Connecting spaces that might otherwise remain detached in some fundamental way, when it comes to information sharing, is essential. Information shapes the evolution of an object.
Transferring information on a cellular level starts with all the possible configurations a particular structure may take. This space exists as a landscape of interconnected microscopic causality.2 An intricate, alien world (alien despite existing inside every one of us), the cellular landscape may be mapped through studying cell-cell communication, what is collectively decided upon then organised biophysically in the object or organism to the point where a structural feature develops. Following this form of information transfer and influential systemic outcome, we come across a signpost reading:
Welcome to Morphospace:
“Raup's (1966) morphospace of coiled shells.” Mapping anatomy and structure from bound possibility spaces (Source: Mitteroecker & Hutteger, 2009, The Concept of Morphospaces in Evolutionary and Developmental Biology: Metaphor and Mathematics)
Using morphospace to understand the role of “evolutionary radiations”—extinction events, or massive shifts in the general conditions—in the developmental cranial structure of dinosaur species over 248 million years ago. Such “events are often associated with exceptional levels of morphological diversification into new habitats and ecological niches, and are termed adaptive radiations”: it seems that new frontiers of morphospace has led to diverse changes in anatomical configuration in the evolutionary space throughout the history of biosphere. (Source: Foth et al, 2021, Rapid Initial Morphospace Expansion and Delayed Morphological Disparity Peak in the First 100 Million Years of the Archosauromorph Evolutionary Radiation)
“Morphospace and evolutionary integration of living and Cretaceous ants.” Morphospace is a bridge between scales of space (e.g. potential anatomical space and biological space, or the space of embedded patterns and real-world causal actions) (Source: Chinese Academy of Sciences Headquarters)
“The morphospace of several groups of reptiles (including dinosaurs) that lived around 230-200 million year ago, using two (undefined) shape characters.” As part of the notion of Extended Evolutionary Synthesis, developmental biology is seen as having directionality when it comes to evolutionary change (Source: Hordijk, 2018, Workshop report: Developmental Biases in Evolution.)
“Morphospace of consciousness” — notice the axis, then notice the area of least activity: the back left quadrant where computational complexity meets freedom of choice and action is thus far free of “artificial agents”. How long will that patch of space remain empty? (Source: Arsiwalla et al, 2023, The Morphospace of Consciousness: Three Kinds of Complexity for Minds and Machines)
Morphospace is an abstract landscape representing all possible structural configurations an organism or system might take, commonly used in biology to discuss the anatomical developments and differences between various organisms. It is now well established that “a morphospace, based on information-theoretic measures, can be a useful construct for comparing biological agents with artificial intelligence (AI) systems” as it can help to “map biological agents such as bacteria, bees, C. elegans, primates and humans; as well as AI technologies such as deep neural networks, multi-agent bots, social robots, Siri and Watson.”3 The scope of an object’s morphospace directly dictates how it develops, inclduing what features it has, and this remains true across biological and artificial spaces alike. Already, some researchers have created a “coevolutionary model integrating niche construction and morphological development to investigate the evolution of adaptive morphology and behavior in artificial creatures within a 3D multi-agent environment.”4 When we ask more “meta” questions like “What is the actual “substance” morphospace?”, without knowing the answer, there seems to be a complex chain of agential assembly at work across various scales.
Indeed, in Arsiwalla et al’s, recent 2023 paper entitled, The Morphospace of Consciousness: Three Kinds of Complexity for Minds and Machines, the “idea that consciousness resides in select regions of a morphospace, that is constructed from function-specific types of complexity” where the “dimensions of our morphospace implicitly entail (or rather subsume) distinct types of intelligence” certainly supports this notion. “Mind”, it seems, extends into spaces outside of the brain and body. Morphospace is not necessarily made of mind, but without a doubt it enables mind to evolve. In such a manner, Arsiwalla et al predict three types of artificial systems corresponding to potential forms of consciousness within future morphospace:
(i) “synthetic consciousness”
(ii) “group consciousness” (similar to “collective intelligence”)
(iii) “simulated consciousness”
Without getting lost in the tall weeds of cognitive science, ever since ideas like “Punctuated Equilibrium (PE)” and the “Red Queen Hypothesis” gained traction in the 70’s, it is a common belief in biology that when organisms undergo vast amounts of morphological change, that is, when their structure has to morph to conform to, in the case of PE, changes to external abiotic (environmental) condition sets, or in the case of the Red Queen, internal biotic drivers, this process can act as a “forcing function” for the emergence of novel and complex features and functions. Drastic shifts, which is the technospheres modus operandi, constantly and continuously morph the shape of agency until consciousness (or at least “cognition”) may come tumbling out of the “bio-digital assembly space”.
From Arsiwalla et al’s “Morphospace of Consciousness” model shown above, notice how, for consciousness to emerge, an agent must exhibit high rates of (i) autonomy, (ii) computation, and (iii) social complexity. Artificial systems score high in computation, less so in autonomy, and low in sociability. But as social networks continue to expand into the digital world research also continues to suggest that “social dynamics could lead to the emergence of more diverse morphology and behavior in [artificial] creatures.”5 And, with 2025 headlines reading:
“Groups of AI agents spontaneously form their own social norms without human help”
Then, to me at least, it is pretty evident artificial ecosystems are becoming less and less human-controlled computational structures and moreso problem spaces in which morphology and evolutionary-style processes naturally emerge and combine to form abstract, very real, forms of agency—across different spatial dimensions!
Back in sunny Glasgow, assembly space— a conceptual space in which objects are incrementally built from simple to complex forms via selection processes—is a core component of Leeroy Cronin and Sara Walker’s AT. Whilst distinct, assembly space appears to pattern-match with morphospace in some fundamental ways.6 Whereas traditional biological morphospace has been seen as a continuous but gradual change shaped by biological constraints, from observing the bio-artificial overlaps we are quickly learning that morphogenesis can become much more of a rapid process than first imagined. Moreover, when engineered effectively morphospace becomes a drawing board for highly adaptive and novel possibilities across scales (e.g. the construction of xenobots and the “reconstruction” of biolelectric pathways guiding cutting-edge limb regeneration). While assembly space focuses more on combinatorial possibilities and stepwise assembly driven by chemical and physical properties, both models are useful for understanding how constraints and possibilities within a semi-structured space influence and shape the evolution of complex objects. Exactly what we need in order to think about artificial “evo-devo”.
When it comes to morphospace, exactly what structure are we talking about here? Well, the short answer is, morphospace is directly influenced by “fitness landscapes”, a conceptual relationship between an organism and its ability to survive in an environment. Back in 1932 the space that causally affects morphogenesis was structured like a polytope—a geometric structure like the Platonic Solids—used because these geometries exist in any general number of dimensions (n) as an n-dimensional structure; in other words, because they can map exceptionally large packets of information. Specifically, these polytopes often took the form of an amplituhedron. Fast forward almost one-hundred years later and not many people remember the polytopic role in the early mapping of morphospace, mainly because nowadays it has since been eclipsed by emerging domains like “Quantum Field Theory” and “Surfaceology”, in which the amplituhedron and the more complex “associahedron” are used to map particle motions and other atomic data. Thus, if we look back at the early accounts of morphospace and notice a pattern-matches to geometric structures now being used to describe particle behaviours, and think that reduce morphospace to the base level of atomic arrangement for new structural “material” to form anatomical features, then the use of the polytope by geneticist Seawall Wright back in the 30’s was prophetic in many more ways than one. Coincidences and anecdotes aside, the ability for an organism to survive in a given space, what we now call its “fitness landscape”, is structured in such a way that it maps points of collective agential activity which then guide the morphological changes. As more and more relational structures bridge the bio-artificial boundary, we find ourselves thinking back to underlying polytopic forms, of the Platonic Space, in which the geometry underpinning the biosphere, which has informed the evolution of biological ecosystems and agency for millions of years, now inevitably and invariantly scales across different spatial dimensions. Only this time it is not from organelle to cell, or organism to society; it is straight into the artificial expanse.
Such an organisational structure can be considered an associahedron in the context of emerging fields like “Surfaceology” and lays the foundations for understanding how well an organism integrates and survives in an environment, as well as how the smallest of causal relations, barely in the physical world, end up assembling a unique yet uniform anatomical structure very much in the physical world. It is the scaling of these processes across spaces that is so interesting with regard to the question of artificial lifeforms. (Source: Mitteroecker & Hutteger, 2009, The Concept of Morphospaces in Evolutionary and Developmental Biology: Metaphor and Mathematics)
Whilst the math goes well over my head, when I think of the most formalised version of “organisational structure” this feels intuitively close—”surfaceology”—mapping the geometries and relations of space at the most fundamental of levels. How well do these processes scale to “larger” spatial dimensions, like the meso-movements of the cosmos?
How could understanding the bridge between spaces of assembly and morphology enable a better understanding of connections between other spaces, say the artificial and the biological?
What kind of complex organisational structures can cross the bridge between assembly → morphology → artificial → biological?
Another world renowned, on-the-verge of a Nobel Prize winning experimental synthetic biologist and cognitive philosopher Michael Levin sees morphogenesis as having a collective intelligence. Yes, probably a lot more collectively intelligent than my organism (I won’t speak for you). But it is not as fair a comparison as it first seems. Morphogenesis—the process driving “action” in morphospace—has access to “multi-scaled competency architectures”—robust problem-solving capacities which span multiple scales of organisational structure (e.g. molecular to whole organism) and integrates what, at the most fundamental level, can be considered “cognition”: goal direction and independent action.7 According to Levin, morphospace shares fundamental homology with cognition, including deep “architectures” of “nested cognition”—hierarchical organisation in which decision-making or information-processing agents are stacked upon, or within one another. Architectural blueprints such as these make agency appear “embedded” at multiple spatial scales. For example:
From the genetic network → cytoskeleton → neural network → whole tissue/organ → whole organism → group of organisms (e.g. swarm) → ecosystem, “cognition” exists across each of these nested, embedded, scales of organisational structure. (Levin, 2022, Technological Approach to Mind Everywhere: An Experimentally-Grounded Framework for Understanding Diverse Bodies and Minds)
Notice how the various types of “competencies” take place across different spatial dimensions (ranging from an “organelle” up to an “organism”), including metabolic, transcriptional, behavioural, physiological and morphospace. Each competency is unique to its spatial dimension, yet together all competencies form an ornate “architecture”, less sprawled and more dispersed, across the entire organisational structure. It is like having a college campus, where each “building” houses a different department, each with their own unique goals and problem spaces, yet each department remains an integral part of the entire campus structure, where the optimal functioning of each individual department sustains the collective goals of the entire “architectural” campus system. (Levin, 2023, Bioelectric networks: the cognitive glue enabling evolutionary scaling from physiology to mind.)
Boolean operations, across natural objects: disassembling objects into constituent parts, breaking them down from largest to smallest to understand “assembly index” (how many “steps” from smallest part to largest) and the total assembly space for the “copy number” (how many of that object exist). If both values are high (if there are many parts between largest to smallest and there are many of those things in existence) then this suggests there is a high degree of “selection” that has taken place in the assembly “history” of that object. There are many stones in the world, but stones have a low assembly index. There are over 8 billion people in the world, a relatively high copy number for the biosphere, and there are many steps between our largest part (especially if we include “consciousness”) and our smallest part (if we denote this to be our most simple starting block of hydrogen or carbon atoms). Thus, according to AT, humans are likely to exhibit complexity—which arguably we do in abundance.
(CCA 3.0)
The “holarchy” has always proved a great visual for understanding the “levelled” and “stacked” nature of interconnected structures across scales, here ambitiously reaching out from the smallest of conceivable subatomic “strings” to the incomprehensibly expansive “multiverse”. If nothing else, a great illustration of scaling “organisational structure”. (Velikovsky, 2014, StoryAlity #119 – The holarchy of StoryAlity Theory)
The original conception of the “holarchy” by Arthur Koestler in his 1967 “The Ghost in the Machine”, which looks at the “whole” organism (holon), its constituent parts (partons), and how they chain together (holon—parton) to form the total living system. Somehow, from this assembly chain, phenomena like “mind” and “consciousness” arises. The question has now become not only how this emerges at the “top” of the stack, but how consciousness, or at least “agency” or “cognition”, is recognised at each distinct level of the unified whole.
(Velikovsky, 2014, StoryAlity #119 – The holarchy of StoryAlity Theory)
“Teleology”, the phenomenon of “purpose” and agency, seems to exist across many different scales of organisation, material, and matter. The selection of objects and parts, from the assembly of miniscule building blocks and the emergence of complex objects, including life, takes place within a hypothetical (yet clearly very real) “assembly space”. Anyone who has used a computer-aided design (CAD) software may intuit such an assembly space, analogous to 3D modelling software like Sketchup or Blender. While CAD intentionally arranges components, in conventional terms assembly space differs crucially by enabling spontaneous, emergent complexity without explicit design. Well, that was before the technosphere expanded. Now computer-aided life threatens to spring up as if from a biobotic Boolean operation.
AT suggests that some of these objects, the ones with high assembly indexes and high copy numbers, feature emergent complexity due to the “selection” of their parts during the assembly process. From part-selection in a combinatorial process we get glimmers of that ephemeral “interconnectedness” at work. Could there be an inherent teleological feature of the assembly space? To “selection”? Already research has suggested “purpose”—goal directedness and independent action—exists across different scales of spatial organisation. If we then turn and think of the artificial space, there is nothing but selection taking place. The entire technosphere is based on positive feedback loops bringing new concepts to life and negative feedback loops directing which features to remove to make way for the next. If AT suggests high selection rates are a precursor to complex object assembly in a certain space, then the hyper-selection driving the technosphere is another high probability nod towards expecting emergent complexity in the objects being assembled in artificial space. But the main point to take away here is the fact that this space of expected complexity is only partially physical; a lot of the evolution, the object “assembly”, is taking place in a digital space we have never lived within. As much as we can enter it, and be adjacent to it, and use it for business or pleasure, to relax or to go to war, we truly do not know what it is like to live inside of any ocean, let alone the digital one.
These nested structures, these points in space and time in which selection takes place, as alluded to in AT, and as suggested by the influence of morphospace in the physical world, may feature agency. Levin has noted how various levels of thermodynamic organisation solves problems in its own space, using some of the same tricks at various degrees of sophistication and scale. At molecular scales, cognition might be metaphorically understood as chemical interactions governed by physical laws, which nonetheless exhibit complexity resembling “decision-making”. From cells, cellular networks, tissues, to entire organisms, extending the idea of “cognition” into morphospace bridges the path back to assembly space. Could it be possible to imagine that each stage at which “selection” takes place within the assembly chain is representative of a form of “cognition” or “agency” at a very abstract dimension of space?
Developing “AI agency” continues to be a top priority for many leading tech companies. True, agency in this sense is different to the agency we see in the biosphere, but it is still early doors. Whatsmore, it is clear that agential markers like morphological features and cognitive functions will not need to navigate the problem space for long in order to find a bridge across the biological scales of organisation into the artificial equivalent; namely because human-based selection means it is already one of our primary technological goals. Furthermore, any cognitive agent inhibiting any level of an ecosystem, across any spatial dimension, deserves attention. Ecology and biology largely take care of that out here in the natural world. Who is looking in the digital niches? When that space is the technosphere, and one possibility is that nested competency architectures and cognition scales into autonomous artificial agency, AT tells us that areas where “selection” is highest may be a great place to keep an eye on for emergent complexity. That is exactly what we are seeing with booms in generative AI and machine learning. Human selection is highest around the objects most “like” us—the ones that talk, predict, edit, sing, do the things people do. What about other corners of the sand box? What about under the sand itself?
By applying Levin's concept of nested cognition to assembly space, in a superficial sense each instance of selection in the assembly process can, in some ways, be seen as having its own form of agency. It is too early to imagine that basic chemical interactions are a fundamental form of decision-making, where molecules “choose” how to interact based on their properties, interpreting signals, and “communication” from other adjacent molecules. But there is evidence to suggest that, as selection takes place along the assembly chain of an object, as more and more parts combine over time, assembly index growing, as more and more of those objects replicate over time and space, the cognitive processes that become evident at the smallest of thermodynamic scales grow in sophistication, developing feedback mechanisms, adopting error correction, and learning to adapt behaviors, until it is scaled to the level of complexity of “consciousness” or “qualia”, such as that exhibited by humans. Where it scales from there, through the biosphere and beyond, is anyone’s guess.
Rapid expansion is a well established aspect of the technosphere. It swells on the consumption of natural resources. It expands into other spheres, changing their dynamics (e.g. an individual “technology stack” being a significant causal influence on a person's life). Through synthetic biology breakthroughs in biobotics like “xenobots”, artificial limb-regeneration, gene editing and “3D-bioprinting”, it is more than evident that morphospace is as much embedded in the techno-artificial assembly space as it is the biophysical world. The full extent is hard to tell. Put it down to one more reminder that the digital expanse should be studied not just as a thriving, biodiverse ecosystem, not simply a place of human-led activity, but a genuine frontier guarding a space in which nobody alive is familiar with. Uncharted territory. An expanse.
“Scales of structure” (e.g. subatomically: proton and neutron arrangement inside a nucleus ↔ thermodynamically: the arrangement of intracellular membranes ↔ classical mechanically: deliberate arrangement of composite materials to reinforce a wall) and “scales of organisation” (e.g. thermodynamically: bioelectric signals passing across the collective cellular network ↔ spiritually: groups of local worshippers gathering on a Sunday ↔ socio-politically: collections of people who obtain a legitimate and effective monopoly of force in a given territory, or a state).
The major difference being the scale of structural organisation (e.g. assembly space relating to “atomic—molecular” and morphospace relating to molecular—organism”)
Levin defines intelligence the same as William James did: as the ability to accomplish the same goals via different means.