Software / Computation

• Architecture ⇔ Building blocks
• Organisation ⇔ Elements (?)
• Operating System ⇔ Data structures
• Theory of evolution ⇔ biochemistry

The Fun Gap: The difference that when you study the system in terms of the computational layer, there is an operating system with a cognitive user in the system that is so much more relevant to the thing you study that they can mess with your experiments just for fun.

Imagine…

Statistical methods that make biochemistry at the thermal rocks until you get something like cells? 🤔

That is not an explanation of life.

What is missing to explain the mind as a theory of knowledge.

Brain software is software and there is no programmer at hand. It is self-assembled software.

• The genes facilitate the growth of self-assembled software on the brain,
• The function of the brain is to allow for software to self-assemble
• The biology of the brain is the theory of brain software, and it is the theory of self-assembling software

It must be an interactive and dynamic computing framework (Alan Kay, JSC Licklider, John McCarthy).

The Meme World

• Neuronal ensembles survive or don't survive.
• The logic of the memetic engine predicts neuron failure rate (skip rate), and intrinsic firing rate.
• This supports what is called Neuronal Assembly Shift
• devices are things that mess with the neurons in substantial ways, like modifying the excitability, making memory lookups, etc.
• if this is initiated from the neurons, then little computer usage pieces are possible, too

• the memetic engine creates a second biological plane, some kind of simulated virtual world, in which many things are possible. and everything possible is also true a little bit.

Memes are information with causal power, they know how to replicate themselves:

• High dimensional, pattern completed symbols 👈 it's fair to call them memes

Hypothesis: Brain software implements a simulated world supporting high-dimensional, biological software entities.

• the meaning of the high dimensional data structures is their effect on the computational system
• roughly: The connections of the neurons they are made from (?)

• The structure and function of memes is their meaning in the computational framework

• the context criticizes the ideas, their creation is akin to conjecture.
• Conjecture and criticism are an explanation of creativity and knowledge [Popper, Deutsch]
• Make it hard for activity to survive [Abeles 1982]
• Allow 'lateral inhibition', in a broad sense (various inhibition systems).

-> Inhibition is criticism and is an essential component of a theory of brain-knowledge

• Conjecture: Neuronal ensembles are active data structures, representing clouds of counterfactual possibilities. (via their halos).
• Ensembles leave their neurons behind,
• Perhaps more like activity amebas on top of the neurons

—

• A Buzsáki librarian module would allocate random high-dimensional sequences, which an associator would imbue with meaning by doing action, using the corollary discharge and the sensors 3.
• for present purposes: the ideas are there first

• Via the basic Hebbian Cell Assembly notion: An idea is either supported or less well supported, given the context

• The environment/context is roughly speaking whatever else is active right now (but neuromodulators, synassembles³, etc.)

If you shit on all ideas evenly, that's a way for the good ideas to survive. That's the basic notion of a threshold device.

• The most complex critic is the rest of cognition. This fits well with Joshua Bach's notion of consciousness as a coherency function.⁴
• whatever you mean, means whether you survive or not, given the rest of the system

• neuronal ensemble anatomy, or the memetic game of the circuits, or their ad-hoc epistemology

Perception is about the stuff that we don't see.

• It's still there even when closing the eyes
• phantom pain
• Joshua Bach: creatures in the dark
• the second tenant of memetics: inhibit your alternatives on every level of meaning

—

• maybe memes are bags of counterfactual expectations
• memes know themselves how to be used

Up to pre-frontal areas, perhaps…

templates (?)

Also called frames, roles, fillers in cognitive modeling.

The notion of a line is allowed to abstract, the color is a dynamic filler (we already required this above).

• they are allowed to be made from the same stuff.
• neuronal ensembles already are context for other neuronal ensembles
• similarly, neuronal ensembles can mean verbs, transformations, and causality structures via their affordance representations

Composability

In order to be useful, this data structure must have dynamic composability. On-the-fly mapping, on-the-fly making sequences.

Ensembles have a temporal structure.

Perhaps Buzsáki's "Neuronal Syntax" can yield good ideas.

There is only one kind of entity in this system, the high-dimensional, replicating, data structure 👈

Their power and versatility come from the available "devices".

The system only needs to deal with a single type. (I.e. everything is closed closed under the fundamental data structure).

The one datatype in the system is a living, biological entity.

Understanding the brain entails understanding simulated worlds.

—

David Deutsch:

The adaption of a replicator is the knowledge that makes it a replicator, given a niche.

• memes are biological entities in a computational framework
• the memes know themselves how to be used
• memes somehow self-assemble to brain software

• memes are strategic

• Micheal Levin: The cognitive light cone of an agent is how far into the future its goals reach
• In order to solve a Rubik's cube, one has to break existing patterns.
• memes are harmonic and they have extended phenotypes [Dawkins 1982]

• Donald Hebb: The Organization of Behavior 1949 - same thing on the level of cell assemblies
• Daniel Dennett: Consciousness Explained 1991 - similar thing on the level of ideas
• Here, it is on the level of software.
• meme compositions are allowed to be made from the same data in HD software.
• meme compositions somehow are allowed to be arbitrarily sophisticated users of the computer they run on
• Competent memes cause the computing framework to reproduce themselves:

• These are computational survival software user entities using an operating that evolves to be useful
• What is useful software? This is a programming design philosophy question. What is expressivity?

• This somehow must make interfaces and languages, "hide the details"
• We know this "hiding" thing is true for sure because we are a cognitive user ourselves.
• Daniel Dennett called this hiding the details the user illusion. I call this the magical interface instead.
• Explaining cognitive user entities was required for a theory of cognition, so it is sweet when the explanation requires them, too.
• Conjecture: Languages cannot use themselves (otherwise it has its own ideas, and isn't a good language), therefore, each language needs a user level. Systems of expressions are required for powerful software design. (philosophy of programming, underdeveloped)⁵, ⁶ (future constructor theory of programming and design?)

• expressivity predicts that the elements are composable
• a language that has it's own ideas isn't a good language
• Imagine you need to use the word banana in every sentence

critical periods, synaptic overproduction and synaptic pruning 👈

1. User-level entities will want their producers to be general
2. Producer level entities want to be general, and composable (building blocks). Because systems of expression are even more powerful and useful.
3. users want to be competent composers of useful languages.
4. user level entities should be allowed to be symbiotic with the languages they use, creating harmonic, expressive building blocks at the producer level that I label systems of expression 👈

• Neuronal Syntax (Buzsáki)

• 'analogy', templates (high dimensional computing helps)
• action, transformation, change (?)

https://lab42.global/arc/

• The constructor theory of knowledge would be an extension of Dawkins's replicator theory
• Chiara Marletto says this would also talk about civilization and technology

A constructor theory of computing

I want to make progress in describing theoretically what computing frameworks, interfaces, languages, and operating systems are.

Machine Learning?

Learning and behaviouristic paradigms are not explanations of the mind.

• Dragons were not real, dinosaurs were real
• Perhaps spirits were not real, but living ideas were
• Living ideas: memes made from stable information that make software entities
• Information represented via high dimensional data structures in a computational system
• … made from neuronal activity
• Intrinsic activity and central pattern generators: Perhaps they are like the spirits, the inner life, the creative spark
• The idea of life in a virtual simulated world, representing the essentials of the outside world. A seed grows into a tree, this tree-growing causal structure can be represented as a tree spirit living in the seed. Those causal structures, internal mechanisms and surface interactions, are real things in the world.

You make an area that is this hard to spread into, for instance with low excitability or frequent resets.

Dan Dennett: Fame in the brain.

Fits with 'global work space theory'.

Perhaps an ensemble can say think me, by spreading a little bit into the challanger node.

The challanger node is 'merely' a handle for using the rest of the system.

• have a perspective mechanism that makes random inputs to neurons
• modifies the attractor landscape dynamically, from within the neurons (perspective-effectors)
• ≅ attention in machine learning

Just an idea, perhaps one aspect of the tip of the tongue.a

• the mechanism for the feeling of tip-of-the-tongue is here in the notion that a vague "narrative unit" coming from user-level entities is undergoing destabilization, "fleshing out", "filling the blanks" This fleshing out fails to provide the memory lookup required, yet the "speak" part is being "fleshed out", too. The hypothesis is that this fleshing out usually is a way for the producer level to find a result, but didn't work.

• neurons don't heal- synaptic 'overproduction' is a misnomer, when it is part of the functioning
• critical periods: Language phonemes

Perhaps brain software is made out of ghosts of living memes, like a landscape carved from rivers.

Kantian pre-fill mode: prepare a net for some physical inputs.

• [developmental neuroscience, sensor motor inputs ]
• traveling waves?

Has an analog in machine learning. [pretrainend convolutions I think].

• Consider the banana language (every sentence is required to have the word banana in it). It is less useful than English.
• The repertoire of a system of expression is the set of sayable things.
• everything possible to say should be sayable.
• You want to be able to compose the elements freely.
• (same as 'the living language problem').

Perhaps dreaming is keeping the system of expression smooth that we use to create our 'waking dream'.

• Reminds me of the book title by Wiener The Human Use of Human Beings (1950).
• Cultures are living entities, just like ideas are
• Systems of ideas use their resource level, perhaps these 2 things together are a definition of technology

The new generation will have no trouble accepting that consciousness is a virtual machine. They are used to virtual machines when using the software of their computers and phones.

From memory, not verbatim Dennett 2023.

• Consciousness is a virtual simulated reality created by software ✓

New questions:

• What is software?
• What is virtual?
• What is reality?

Pentti Kanerva. Hyperdimensional computing: An introduction to computing in distributed representation with high-dimensional random vectors. Cognitive Computation, 1(2):139–159, 2009. https://doi.org/10.1007/s12559-009-9009-8.

Rafael Yuste Lectures In Neuroscience, 2023

Buzsáki, G. (2019). The Brain From Inside Out. New York: Oxford University.