Do you think fractals (i.e. iterative and self-similarity) are weird? Well, it isn’t as weird as biological iterative processes.
The Alien Style of Deep Learning Generative Design
What’s even weirder is that humans have an intuition that something appears organic. What does it actually mean to have an organic design?
Christopher Alexander, an architect, who wrote ‘A Pattern Language’ that has immensely influenced software development, wrote four books exploring this idea (see: Nature of Order).
Organic design and its biological underpinnings are extremely complex. Allow me however to focus on a narrower scope, something that is less complex. General intelligence is something less complex and something that is also organic in nature.
Darwin’s theory of evolution appending the existing orthodoxy that the species that inhabited the world was a fixed thing. What society has yet to come to grips with is that our brains are *not* fixed things.
Brains are ‘livewired’ to their sensors and their bodies. They learn to interact with this world by being embedded in the constraints of this world. All learning is entangled in context and all meaning of language is also entangled in context.
We however create explanations of this world through concepts that are disentangled from their context. From a Peircian perspective, our signs evolve from icons to indexes to symbols. This sign evolution leads to a loss of information.
Humans are still able to understand each other through language (i.e. a sequence of symbols) because our interpretations annotate words with meaning. But in all cases, it is our subjective meaning of the words.
Subjective implies that it interpretative relative to the interpreter’s reference frame and hence imagined context. This is Wittgenstein’s Picture theory of language interpretation.
Dall-E and Wittgenstein’s Picture Theory of Meaning
What then does it mean when we say that brains are constructed from the inside out in an organic manner? How does this inform our understanding of general intelligence? Why is it different from the brain as a computer metaphor?
I’ve already explained why the brain as a computer is a horrible metaphor: medium.com/intuitionmachi… So let me skip that question.
The Brain is like a Computer is a Terrible Metaphor
I’ve already explained why organic design is different from engineered design. So let me skip that question too!
How Biological Design Differs from Human Design
So let’s focus on how general intelligence is informed by organic design. We’ve already established the error-correcting nature of both organic design and cognitive development:
Error-Correction is the Key to Understanding Intelligence and Consciousness
Christopher Alexander uses error-correction to explain why towns that emerge organically look very different from those that are architected. The activity of living modifies the world in a gradual manner. Adjusting the world to make convenient the pursuit of life.
Alexander proposes 15 recurring patterns that he’s observed in organic design:
However, these concern themselves with physical structures. But do the mental models that we grow and cultivate in our minds also follow the same organic principles?
I conjecture how our minds create the kind of thinking required for the left hemisphere is based on different prioritization of inductive biases as compared to the right hemisphere.
I suspect there is a misconception that system 1 (intuitive) is mapped to the right brain and system 2 (deliberate) is mapped to the left brain.
The left brain is livewired to be competent in sequential thought while the right brain is livewired to be competent in parallel thought. One is egocentric and the other is allocentric. One is symbolic and the other is empathic.
One is reductionist and the other is holistic. One is noun-centric and the other is verb-centric. One emphasizes individuality and the other the collective. We can make many analogies about the dichotomy between the two hemispheres of the brain.
However, the analogy that system 1 resides in the left brain and system 2 on the right is incorrect. Both systems are system 1. System 2 is just a reflective instantiation of system 1. The brain always employs parallel mechanisms for its processing.
Any notion of sequential circuitry is an illusion. It is a similar illusion that we describe as consciousness.
We however do not lose our consciousness when our left brain is damaged. Indeed we are incapacitated in our ability to perform sequential action or understand language, but we do remain conscious.
Consciousness like system 2 implies the capability of being reflective of one’s thoughts. The kind of reflectiveness in the right hemisphere is of the allocentric kind. The kind that is in the third person.
One can argue that Robert Kegan’s model for adult development cannot happen without the reflectiveness of the right hemisphere. Adult development is a consequence of the egocentric-self expanding itself towards the allocentric-self.
The two hemispheres are livewired for different concerns. The fluidity of human thought is a consequence of the the interplay of these two. We would not be able to solve Bongard problems without the interaction of these two hemispheres.
A Bongard problem is a rule finding game. The objective is to find the one consistent rule that is self-consistent for the graphics on the left and the right. Furthermore, the left rule is the anti-thesis of the right rule.
The left hemisphere will be very good at identifying patterns. Unfortunately, it will be incapable of find alternative patterns. It requires the right hemisphere to perform a search for alternatives.
Our system 2 reflective and deliberative process kicks in when we are validating our solutions.
But analogous to AlphaZero that makes system 1 judgments about policy and value, the interplay between the left and right hemispheres makes judgements about patterns and alternative patterns.
That is conceptual wholes and conceptual distinctions. That is symmetry and symmetry breakings.
When we attempt to solve a Bongard problem, we spontaneously create alternative ways of matching patterns. One sometimes hears the words ‘inductive bias’ to refer to this. Unfortunately, the vocabulary we employ to describe patterns is impoverished.
The starting point for a vocabulary of ‘inductive biases’ can be found in Alexander’s 15 patterns.