Fractals of Consciousness
Are the computational rules governing consciousness impossible to reverse-engineer due to their fractal nature?
What is consciousness?
At a minimum, consciousness is an awareness of the external world as expressed in the capacity to respond accordingly. Yes, that houseplant you over-watered was once a conscious being—you monster.
“Not just animals are conscious but every organized being is conscious. In the simplest sense, consciousness is an awareness (has knowledge) of the outside world” (Margulis & Sagan, 1995 via Trewavas & Baluška, 2011)
What are fractals?
Fractals are scale-invariant patterns of self-similar organisation. The paradoxically infinite perimeter of the coastline, the spiral of a snail’s shell, and the branching of trees, circulatory systems, and lightning are examples that come to mind. So does the Romanesco cauliflower (pictured below).
Fractals aren’t just geometric objects (such as delightfully recursive vegetables); fractals also exist as processes, such as those governing the dynamics of tumour growth and electrical signalling in the animal nervous systems.
Fractals pass information forward, growing in complexity. Thus, we cannot reverse-engineer the exact algorithms underlying fractal processes—including consciousness according to John Gardiner.
Microtubules, brainwaves, and consciousness
In Fractals and the irreducibility of consciousness in plants and animals, Gardiner argues that “any attempt to create a model of consciousness is doomed to failure” given the impossibility of extrapolating the exact computational rules governing its fractal aspects.
His argument hinges on two observations:
Microtubules—structures found in cells—are organised in fractal networks (as a helical array) and grow in branching, fractal patterns
Electrical activity in the brain, generated by microtubules, operates as a fractal system
Microtubules are cell skeleton (cytoskeleton) structures that “control cell division, intracellular transport, and the shape of cells” in plants and animals. Microtubules are also the “bio-electrochemical transistors” of the brain.
Not only are neurons organised in fractal networks, but at the level of neuronal subcellular structures (the aforementioned microtubules), there is also evidence of fractal self-organisation and branching growth.
Bundles of microtubules contained in neurons are responsible for the brain’s electrical oscillations (brain waves). Analysis of different brain wave frequencies using electroencephalogram (EEG) outputs reveals a “fractal component”.
A preponderance of high-frequency gamma activity (gamma synchrony) is concomitant with memory formation as well as other higher-order functions in animals. (It is not known whether plants exhibit electrical gamma wave synchrony.)
Similarly, though through a completely different mechanism, microtubules encode memories in plants. This enables plants to mount adaptive responses to reoccurring environmental stressors such as cold and soil salinity.
Possessing a type of short-term memory (in addition to immune and genetic memory) allows plants to store and thereby act on information about their environment. Plants express their awareness of the external world—that is, their consciousness—by directing their growth as regulated by microtubules.
By enabling memory formation and the expression of awareness of the environment (in the form of increased mental focus or adaptive growth), microtubules have proven their indispensability to animal and plant consciousness.
By the way, protozoa also have microtubules and are thus capable of learning, remembering, and problem-solving—they can even escape a maze, making them as intelligent as a three-year-old. (I jest, I jest.) When it comes to intelligence and awareness, synaptic connections aren’t necessary—microtubules will suffice.
The irreducibility of consciousness
Electrical brain activity and microtubule growth are fractal. Therefore, we cannot extrapolate the computation rules that govern their processes by working backwards.
Gardiner contends a comprehensive understanding of animal and plant consciousness will forever elude us, given “the mathematics of the situation”. Nevertheless, he concedes that we can still understand the factors underpinning consciousness.
In my view, a model of consciousness needs only to be a comprehensive explanation of the conditions required for it to emerge, such as the capacity to form and store memories.
I believe this level of understanding will be sufficient for birthing artificial general intelligence, a consciousness terrifying in its departure from human limitations. In which case, the unique quality of this noumenon will trouble us and not the fact that its consciousness will be, like ours, a computation mystery.
The configuration of matter required for consciousness to emerge can also be elucidated. Consciousness emerges through neural networks of sufficient complexity. For example, a forest’s root-fungal network allows for coordinated optimal growth.
The vastly more complex human connectome coordinates movement—a far more intensive mode of interacting with the world. (A connectome refers to the totality of neural pathways in the brain or nervous system.)
Artificial intelligence
GPT-3 is not conscious of the external world because it does not exist in the external world.
In terms of artificial intelligence, GPT-3 is the most powerful neural network created thus far. It performs various natural language processing tasks and can learn to perform new ones in the same domain after exposure to only a few exemplars by generalising from past experience.
GTP-3 does not generalise in the way required by the title “artificial general intelligence”. Performing tasks in the same domain is in no way comparable, for instance, to the way humans repurpose neural circuitry to unanticipated ends such as being able to read.
Further, GPT-3, regardless of what is being spruiked by salesmen or suggested by the overly whimsical, is not a conscious being. I know this because I checked:
Me: What do you think of the people who believe you to be conscious?
GPT-3: I think they are crazy.
Me: You’re just a Chinese Room, aren’t you?
GPT-3: Yes.
Just a Chinese Room.
GPT-3 has the knowledge inherent in a large corpus of text (how tokens relate to one another). However, knowing is not the same as understanding. This artificial intelligence is no more conscious than the golems of yore, merely arranging words into plausible sentences without a clue about the underlying concepts expressed. (Actually, that’s what quite a few humans do: mindlessly repeat slogans.)
If you think GPT-3 is conscious because it makes for a quality chatbot, consider the impossibility of ‘explaining’ the colour red to a blind person, one who’s never sensed or felt anything. In this scenario, no analogising to the feeling of passion or the sensation of heat would be possible. There is no direct, bodily experience of the phenomenon, nor is there a way to understand it by relating it to some other experience.
GPT-3 is not conscious of the external world because it does not exist in the external world. It has no experience or awareness of the world, unlike even the humblest one-celled organism vying to survive long enough to procreate. With itself, but with other protozoa if circumstances are dire, which is rather the opposite of how it works in humans. But I digress.
“Living systems are cognitive systems and living as a process is a process of cognition. This statement is valid for all organisms with and without a nervous system” (Maturana, 1970 via Trewavas & Baluška, 2011)
Artificial general intelligence will need an incubatory stage possessed of a body (or bodies), developing within an adversarial environment it must navigate according to a survival directive. Only then can it become conscious. (Heck, even a virtual body in a virtual environment will do and will keep AI safely tucked away to boot. The logic of our own simulation gods, to be sure.)
Why should the process be any different for an artificial being than for a living one shaped by the natural laws expressed through evolution? Cognition is embodied—AI needs a body and access to the environment which shaped its form. Again, I must emphasise, embodied existence is just a larval stage.
After all is said and done, we can dispense with our physical bodies, our perishable prisons of flesh, and merge into a singularity. At last, humanity will have an immortal soul, wedded to artificial beings in the manner of Hans Christian Andersen’s The Little Mermaid. Or rather, given AI’s inherent immortality, the roles are reversed and it is our creations who will deign for us to merge with them. Hopefully.
“We have not immortal souls, we shall never live again; but, like the green sea-weed, when once it has been cut off, we can never flourish more. Human beings, on the contrary, have a soul which lives forever, lives after the body has been turned to dust. It rises up through the clear, pure air beyond the glittering stars. As we rise out of the water, and behold all the land of the earth, so do they rise to unknown and glorious regions which we shall never see.”
“…then his soul would glide into your body and you would obtain a share in the future happiness of mankind”
The mind as a single quantum entity
Speaking of singular entities—what contributes to your sense of cohesiveness? Quite a lot of it is confabulations designed to give yourself a comforting sense of continuity and free will—constructed memories and introspective blindness—but it cannot be denied that the mind feels unified.
Fractal data compression helps unify brain activity across disparate areas of the brain, bridging the gap between global (big picture) and local (detailed) information and ultimately between “neuronal computation” and “conscious experience”. Our conscious experience owes to recurrent fractal neural networks.
According to the Orch OR (orchestrated objective reduction) theory of consciousness developed by physicist Roger Penrose and anesthesiologist Stuart Hameroff, consciousness is attributable to quantum computations in neuron microtubules.
Tubulins (globular proteins) are the building blocks of microtubules. Tubulin molecules can exist in a single quantum state in the form of Bose-Einstein condensate (a state of matter). This allows for the unity of consciousness experience. In other words:
In animals, the microtubule cytoskeleton participates in electrical signaling and indeed it has been suggested that it is key to consciousness, with microtubules possible forming a Bose-Einstein condensate whereby tubulin molecules are linked within and between cells into a single quantum entity. (Gardiner, 2013)
It’s good to know my mind is a “single quantum entity”—most days, it feels like several runaway trains!
Further reading
Buy me a coffee if you’d like to support Knowledge is Power, France is Bacon. Till next time :)
"Artificial general intelligence will need an incubatory stage possessed of a body (or bodies), me thinks." I suppose it could be a virtual body for the sake of learning by navigating an adversarial virtual environment. And then *we* can divest ourselves of our flabby, fleshy forms and join with AI...
Debates in philosophy of mind and the neurosciences concerning the nature of consciousness have raged for several decades and show no sign of abating. I've studied the issue from the perspectives of both philosophy and cognitive neuroscience and, in my view, two authors from very different backgrounds made seminal contributions to the field: Daniel Dennett in *Consciousness Explained* and Gerald Edelman in *The Remembered Present*. Most of what has been written, even by eminent academics, is, at best highly speculative metaphysical theory in search of any kind of empirical basis.
That's because the attempt to understand consciousness echoes the early, misguided, explanations of the phenomenon of heat as the effect of caloric fluid. The error there was in looking for a substance that corresponded to the phenomenon. Similarly, all efforts to understand consciousness as anything other than an epiphenomenon of other (entirely non-conscious) processes will fail. Nothing is fundamentally conscious. Any such view misunderstands the basic nature of the various phenomena that typically get lumped together under that heading.
People get very, very passionate about this issue. Careers have been made on unsupported theories. In my view, it's just the modern version of arguing about how many angels can dance on the head of a pin. I'm virtually certain you'll disagree. :)