Introduction; the late, great, human brain

 

Text © Sean O Nuallain

Code © Sean O Nuallain, David Bernal, Mia Nakamura and Stan Klein. Released under GPL

There is as yet no consensus on where consciousness fits in modern science. By the end of this course I hope you will be convinced we have to take into account the Arts and other parts of human existence. However there is much that we can do within the purely scientific context.

 In particular we can approach brain science in a responsible and creative way. As you will see with just a few simple simulations which you will be able to run yourselves we can achieve extraordinary results.

Well into the second decade of the 21st century Karl Pribram and his student Walter Freeman the third, who were both members of the great generation, were teaching us how to study the brain. The approach in general can be called neural dynamics but goes well beyond what we currently understand by that term. Literally tens of billions of euro and dollars have been spent on fruitless quests using other approaches since 2000 and it is now time to give these two great scientists a chance to speak to us.

Karl Pribram worked with our team in Dublin in 1999. He had been attending a conference that we organised and one of of my computing PhD students took him out to see Dublin over a few days. The result was the only computational implementation of his work during his lifetime. That will be the content of our first few sessions and you will learn how to to run some simple programmes implementing his vision.

 In 2007 I took over running Walter's lab at UC Berkeley. That began a close Association which continued until he passed away in 2016. He was an enormous supporter of my group foundations of mind and published both with me and group several times. We have several videos of the talks that he gave us and most importantly we have code you will also be able to run.

Pribram will be our first port of call. He was a fascinating character and his range of reference ran to David Bohm and physics as well as the nitty-gritty of neurosurgery which he performed before committing to pure research. His most famous speculation is that the brain worked like a hologram. In fact the whole cosmos was a hologram and we in some sense were a reference beam that unveiled the reality or implicate order underlying all things

. So what we experience is an explicate order as both he and David Bohm asserted. We will return to such speculation. The major point to derive from this is our sensory systems analyse a noisy, infinitely large set of sensory data, and produce compressions that are useful to the cortex. Of course, this is pretty much exactly what AI systems are doing at the moment but the brain and sensory systems in humans do it infinitely better. 

Let us call the initial sensory data the sensory plenum. It is essentially a mess of frequencies that we must somehow interpret to derive information and meaning from. Karl believed that the mathematical techniques used in the brain were what is called the fourier transform and the Gabor transform. There is a good explanation of the former here and you will need it for practically everything else you do in engineering and indeed the modern world in any case

https://betterexplained.com/articles/an-interactive-guide-to-the-fourier-transform/

The central idea - which seemed astoundingly radical when when it was proposed by one of Napoleon's scientists two centuries ago - is that every complex of frequencies can be decomposed into sines and cosines. This was not at all obvious to put it mildly! And what we are about to show you how to do is how to get neurones to do this! We are first of all going to have to undermine the indoctrination you may have undergone about neurons in a brain science class.

In the 1950s the English scientists Hodgkin and Huxley did some remarkable experimentation on on the giant neurones of squid.

https://en.m.wikipedia.org/wiki/Squid_giant_axon

This allowed them to to formulate a detailed mathematical model of neurons in general. Several of us noticed that the neuron model was in effect a resonator. In fact you could model it using the idea of a simple harmonic oscillator - and a pendulum is about the most simple harmonic oscillator you can get ;

https://simple.m.wikipedia.org/wiki/Simple_harmonic_motion

Your brain science indoctrination told you that neurons received what are called afferent signals and add them up up to see if they exceed a threshold. If they do then an efferent pulse emerges from the neuron. Essentially all of this is is premature closure which is a polite academic way for saying complete bulshit

What happens instead is that subthreshold oscillations which are oscillations of the electrical charge of the membrane of the neurone occur continually a maximum rate of about 200 cycles per second. If a massive signal from every single other neuron in the brain arrives at wrong time during the cycle you don't get firing. If even an inhibitory signal from another single neurone arrives at the right time you do get firing. We are going to give you a set of diagrams of how this works later on.

To return to academic speak, the integrate-and-fire paradigm is a limit case of the more encompassing resonate and fire paradigm we are about to teach you. By the end of this class you will have run a few simple programs showing how it works. But let's return to the sensory plenum

There is famous work by Hubel and Wiesel showing the brain has blob detectors and edge detectors and so on. There is similar work also Nobel Prize winning by the Moser lab in in Norway. We are going to throw that beautiful phrase premature closure at both of these Nobel winning projects. We are going to say that Karl Pribram was right and that they are wrong as they assume way too much about the sensory plenum and its structure. But first we are going to take a little detour

Norway is not a member of the European Union but this Nobel Prize winning work was funded by the taxpayers of the European Union. In 2009 a South African called Henry Markram announced at a TED talk that he could in fact solve all the problems of the brain in 10 years

https://www.ted.com/talks/henry_markram_a_brain_in_a_supercomputer?language=en

The bureaucrats in the European Union could not resist his charm and gave him a billion euro to realise his project in Switzerland , which is similarly not a member of the European Union. The project fell apart in record time and in 2014 a petition featuring over 800 scientists forced the project closure in its contemporary form. This is but one of many such projects which have ended in disaster since Walter and Karl were in their prime.

So what is wrong with all of these projects? With the Henry Markham project essentially everything from management to the central paradigm. For Hubel and Wiesel and the Moser project it is a lot more subtle.

Karl as a card-carrying mystic viewed the sensory plenum as something transcendent and indeed sacred. It could not be processed with blobs, angles, or grid cells. We are going to have to be able to take any random set of frequencies and analyse them into the type of components we can use to communicate to higher areas of the brain. Fortunately that is exactly what the fourier transform does. So does the other transform invented by Karl s fellow Hungarian Gabor.

So we started with the neurone as a harmonic oscillator. Simply take the following code and dump it into to the free resource octave.

https://octave-online.net/

Here is the code

clear all; close all; clc;

neur=[10 0 0] %Getting things started

synapse=[.95 .3 0; -.3 .95 0; 0 0 0]

for t=2:1000; %get it close to zero at end of run

neur(t,:)= neur(t-1,:)*synapse ;

end

plot(neur(:,1))

Congratulations! You have just run your first neuroscience simulation! On the x-axia you vae time. On the y-axia is the amplitude of the oscillation which decays over time.

Don't worry for the moment about what all the symbols mean, but just enjoy the fact that we have what is called a forced harmonic oscillator with damping. And it is this damping that allows us to simulate very readily the depressed period after the neurone fires when it is is effectively inhibited from firing again for a short time

We began to realise that this could be an entirely alternative dynamics for how the brain works. We could have two neurones together which are connected and if you throw this into octave you will see what that looks like

clear all; close all; clc;

ampAll=[10 0 5 0]; %initial

N=4; %2*number of oscillators

dt=.01; %each sample is 0.1 sec

Tmax=10;

freq=2*pi; %radians per sec

Nsamp=Tmax/dt %1000 samples is

neur(:,1)=ampAll';

HarmOsc=[ 0 1; -freq^2 -.5];

synapse=[HarmOsc zeros(2,2);zeros(2,2) HarmOsc]

for it=2:Nsamp; %do calculations every del_t=0.1 sec

neur(:,it)= neur(:,it-1) + synapse*neur(:,it-1)*dt;

end

plot(dt:dt:Tmax, neur(1,:),'b'); hold on

plot(dt:dt:Tmax, neur(3,:),'r')

Of course, it would be great if we could all afford Matlab, in whcu these are originally written. Again, y-axis is amplitude and you can see the two neurons'' amplitude dacays over time marked on the x axis. If you continue on the course you will have code for three connected

neurones later on and watch what happens when one of them fires. You are not expected to understand the code for the moment but please be aware that we have implemented all of this and how exciting it was to implement finally Karl's holographic Vision

So we are saying that these Nobel Prize winners Hubel,  Wiesel and the Mosers are just assuming way too much about structure in the external environment. We are not saying they are necessarily wrong but just incomplete. Ok the Moser grid cell project did not mention the problem that Edward admitted in conversation that grid cells do not develop if the rats grow up in a circular environment. So maybe they are wrong?

Not really our business as we're trying to do justice to the insight behind the holographic vision. Specifically what we're trying to do is to ground the holographic vision in the details of sensory processing. So far we have the idea that neurons resonate with subthreshold oscillations of the membrane potential, the elctrical c harge on the « skin » of the neuron. and from that we can develop neural nets of a much more biologically realistic nature than anything so called deep learning uses.

But how do they do sensory analysis? In particular how do they implement the fourier transform? Because if they can implement the fourier transform  using the Architecture of the basic neurons it's also possible that they can do quantum computation ! The first algorithm that worked in quantum computing was based on that transform.

Picture this then. At birth there is a tremendous mortality rate of neurons which do not get to recognise anything in the environment. This has been called neural darwinism. The rest of the Neurons co-operate together in order to recognise things in the environment and process them to produce a Fourier transform which is useful to the cortex. We found an algorithm that could do this which essentially adjusted the delay of signals arriving from other neurons. That proved to be sufficient to give the neurons the functionality they needed without blobs, angles. or grid cells. We had a hopeful beginning of the holographic view of mind

Walter Freeman III

The career of perhaps the greatest neuro scientist in history got off to a very shaky start. For a start his father was the notorious lobotomist of the same name. Among the approximately 4000 lobotomies that he did was a botched operation on rosemary the sister of president John Kennedy. Although she lived on for more than 7 decades it was us a quasi vegetable and it says a lot for the power of the Freeman family that he got away with it.

This lecture is termed as it is becuase, whatever about the claims of graphene in the « Vaccines », several scores of millions of PPE had to be withdrawn becuse of graphene therein, clearly a dosage test to chack how mucj of this substance can be tolerated . Graphene is the favoured substance for controlling neural firing in the brain  from outside it. You who have read to here know this neural firing is a bull in the china shop which is the late, great, human brain.

Not surprisingly Walter the third had problems choosing a career. As somebody who worked in his laboratory I can testify to his technical genius even well into his 80s. The navy was delighted to have him enroll at the age of 17 in 1944 to help them with new technologies like radar. There is a photo of him from that period in which he is clearly happy.

But on his release into civilian life his career problems began again. He was refused admission into Stanford and began years of academic wandering between electronic engineering, philosophy, and much else before finally following a family tradition and enrolling in Yale - without a prior degree - to study medicine.

During the 1950s drug treatments were beginning to take over from lobotomy as a therapy for the range of tragedy and consequent maladjustment that is bizarrely termed mental illness. Seeing the future, his father told Walter perhaps he should use the mathematical and other engineering techniques he had deployed in the Navy to study the brain. And so began one of the great academic careers in history which revealed many of nature's secrets.

In 1959 he began work at UC Berkeley which was his home for the next 57 years. He initially used many of the techniques he had perfected during his Navy years and not surprisingly the Navy funded him for a decade while the institutes of mental health did so for four decades. The terminology which can be found as far back as 1959 included terms that were then new to neuroscience, like state variables, which was a formalization of intuitive notion about characterising a process by one of its attributes. It would also have been surprising to find the use of the French mathematician Laplace which one can find in his early papers.

Walter summarised this innovative work in a 1975 collection « mass action in the nervous system ». He then decided it was an oversimplification. He was going to have to look as the new science of dynamical systems and the possibility of of chaotic interactions in the the brain and particularly in the cortex. Far from following his father into the lobotomy trade he once had a dream in which all of the cats he had used as subjects appeared to haunt him! From that point on the only experimental subjects used were rabbits who had electrodes to detect the eeg implanted under anaesthesia with veterinary supervision. On occasion, human subjects undergoing surgery for epilipsy volunteered data. We use perfectly adequate simulations based on these data.

But Walter the philosopher was not dead. Far from it! He took classes in phenomenological philosophy from Hubert Dreyfus at Berkeley and then wanted further across campus to Tolman Hall learn the elements of behaviorism. Few people have made more use of the the small size of the main UC Berkeley campus to educate themselves than Walter. In the unlikely event of a similar Promethean figure accumulating the number of data points that he did, they will have to live another lifetime as well to emulate Walter's expertise in merleau-ponty, Thomas Aquinas, behaviourism, and physics.

The view of brain and cognition that emerged is remarkable and radical. First of all there are no representations in the brain. What happens instead is that, over a chaotic substrate, the brain maintains a set of perception-action templates, all of which are characterized by the way that the substrate has its amplitude modulated for that particular template. So you see a green traffic light and you'll go but you'll see a green sofa and you stop and sit down.

Therefore the templates are not related to individual stimuli. They are related to the meaning of the stimulus for the individual that is in particular how the individual will behave. « Holonomic » is the keyword here in that when one template changes the entire set of templates change. This really should have been sufficient to gain Walter a Nobel Prize and one was in fact given for this basic insight due originally to him.

But that is not all. Far from it! So stimuli evoke reactions in the senses which then transit over to the part of the cortex dealing with that particular sense. What Walter found is that, in the transition to the cortex, all the details of the original stimulus are lost and all that remains is the amplitude modulation. This he termed the brain laundry!

Therefore the perception-action cycle according to Walter Freeman, who noted the similarity of his schema to Maurice merleau-ponty, as to Thomas Aquinas, went as follows. A stimulus provokes a reaction in the senses and this is is interrelated to to what he called an attractor landscape. For example one such attractor would say to go because it's a green light another to stay because it's a green sofa. Whichever was the most apropos the stimulus is the one that gets selected and that creates a massive avalanche of neuronal action in the brain. Eventually this reaches parts of the higher cortex and we get a subjective feeling.

He was a remarkably sophisticated philosopher for somebody who was primarily a scientist. With respect to the relationship between mind and world, he made the point that as humans all that is required is consensus on what symbols mean. There does not need to be any denotation. In actual fact, at the sensorimotor level you could characterize the relationship between mind and world in strict numerical terms using a quantity called pragmatic information and concepts from behaviourism and from thermodynamics.

Moreover his templates which operated on the verge of chaos were stabilized by interaction with the environment. This is is an exemplification of the schema worked out by Jean Piaget in psychology merleau-ponty in philosophy as well as Rodney Brooks in robotics and much else.

But there was always more! So he described the first half a second after the presentation of an afferent stimulus as three separate avalanches in the brain culminating in in some subjective experience. In the process of each avalanche a noisy gas like state became converted into a liquid through a phase transition. The brain did not care about the cost in terms of energy as long as spatiotemporal variation was catered for in its neural system which is incredibly energetically expensive.

In later life he began to couch this process in terms of quantum field theory. There are many advantages here like the fact that memory can be regarded as a compilation of ground States which require no energy. We will leave this until later.

If you wish to enroll on this course ;

1. Run the two snppets of code above

2. Do a screenshot of the two diagrams

3. Send them to universityofireland@protonmail.com by sept 5 2021 at which point we will start formally and in close interaction with everybody enrolled