Tony Lawson vs John Searle on Money: Why Lawson is Right - Money is "Positioned Bank Debt"

Tony Lawson has presented a fascinating argument that money is symbolically codified central bank "debt" in a paper from a couple of years ago (see https://academic.oup.com/cje/article-abstract/42/4/1165/5054022) John Searle, who had a significant intellectual engagement with Lawson prior to his dismissal from Berkeley for sexual harassment (see https://www.dailycal.org/2019/07/02/former-professor-john-searle-loses-emeritus-status-over-violation-of-sexual-harassment-retaliation-policies/), objected to Lawson's theory as being "incredulous", arguing that his own theory of "status functions" with regard to money was correct. Money is real, according to Searle, because a community upholds (trusts) the "status function declaration" that "I promise to pay the bearer" which is made by a central bank. So when the banks lends me money, the obligation is on me to pay back the "debt" to the bank. And of course, that is how we are all taught to think about money.

Lawson makes a radical proposal based on a historical analysis of money. The money that the bank lends is effectively an IOU from the bank to us. Now how could that be? It goes back, according to Lawson, to the goldsmith's issue of receipts for deposited gold in the 17th century. Basically, the goldsmiths offered a depository service where merchants could deposit their gold, and the goldsmiths would issue a receipt for that gold. The receipt was effectively a certificate of debt to the merchant. These receipts became symbolically codified as representing the deposited gold with the goldsmith, and soon the actual presence of the deposited gold was assumed to the extent that it was the receipts that were exchanged without needing to check on the actual gold that was deposited.

It wasn't long after this that the goldsmiths realised that since it was the receipts that had exchange value, they could issue receipts guaranteed by gold that wasn't deposited. Providing not everybody demanded their gold back at the same time, the goldsmiths could honour the value of the receipts that they issued. The receipts remained symbolic tokens of debt by the goldsmith, and complex social relations between bankers, lawyers, borrowers, government, central and commercial banks emerged.

Interestingly, Lawson describes the difference between cash and the electronic representations of money that we are all so used to. He points out that it would be very unlikely for today's multi-millionaires to demand being paid in cash. Cash is the symbolic codification of central bank debt, while commercial banks generate IOUs to the public in the form of electronic records. When somebody withdraws cash they are converting the electronic IOUs from the commercial bank into IOUs to the public from the central bank.

Lawson argues that this is an incredible story (so at least on this point, Searle is right), but it is nevertheless true, and it is so because money is effectively a kind of "technology" which acquires its own perverse logic over history. He cites the development of the QWERTY keyboard as another example - what technology theorists might call "lock-in".

There are far deeper implications for Lawson's theory. What he is basically arguing is that the nature of the social world, including the nature of money, cannot be separated from history. Historical processes are woven into social ontology in the way that cellular evolution absorbs previous levels of evolution (at least according to endosymbiotic theory). There is deception all along the way (Lawson calls it "fraud") - but in the natural world it is the same. mimicry, camouflage, etc all create deceptions which steer the course of evolutionary history.

This also highlights what is wrong with Searle's position. I met Searle twice and found him highly charismatic but somewhat cruel. While not wanting to infer any ad-hominem assault on his intellectual position (which I have gained a lot from and written about here: https://jime.open.ac.uk/articles/10.5334/jime.398/), there was something missing (which I wrote about here: https://dailyimprovisation.blogspot.com/2014/05/pianos-consciousness-and-john-searles.html and https://dailyimprovisation.blogspot.com/2015/06/why-dont-dogs-have-universities-searles.html). Searle's ontology of status functions is fascinating and flat. It is basically a cybernetic theory where what exists in the world exists through the interactions of actors (rather like Pask's theory - see https://dailyimprovisation.blogspot.com/2016/11/conversation-and-contingency-some.html).

I originally thought that Searle was positivist in asserting that status functions assert the existence of things. It seemed to be that they were better thought of as characterising the scarcity of things. It was the other side of the distinction that mattered. Now I would say that the process of maintaining the distinction about the scarcity or presence of anything must account for its own history and its future. That is to say, no stable distinction can be created without an anticipatory system capable or refining the social positions, speech acts, institutional structures, technological resources, etc, in order to survive in an ever-changing environment. Lawson doesn't quite put it like this, but I think it's what he means. Searle, by contrast, has no history and no future. It suffers exactly the same problems as the two-dimensional information view that has led us to Dominic Cummings (see https://dailyimprovisation.blogspot.com/2020/02/from-radical-constructivism-to-dominic.html). It is not the status or even the scarcity of money that is constructed; it is Nothing.

History and the future are the third dimension in the game of establishing trust, and that in turn contributes to the process of constructing nothing which makes possible anticipation. Only with a third dimension is it even possible to create trust and to anticipate a future. The "positions" that Lawson talks about are really multiple levels of anticipation, each with its own history, and built up over a period of time.

What this begins to look like is an evolutionary biological approach where emergence is seen as a fractal process of interconnected anticipations. It's very similar to John Torday's cellular communication theory (see https://www.thethirdwayofevolution.com/people/view/john-s.-torday), and it has many similarities to theories of technology by Simondon, Stiegler, Yuk Hui, Erich Horl,  and others.

There's more to it, but getting money "right" is an essential element in the process of seeing education right.  

Levels of Ability and "Gradus ad Parnassum": a Pedagogy of Constructing Nothing

In education, levels are everywhere. There are levels of skill, stages of accomplishment, grades, competencies and so on. Arguments rage as to whether levels are "real" or not. But obviously there is a difference between someone at Level 1, and someone at Level 8 (for example). Throughout the history of education, attempts have been made to create pedagogies which follow a staged approach to the acquisition of skill. The formalities of these approaches, and arguments about the true nature of levels (for example, whether one might naturally acquire high levels of skill without the formalities of a levelled pedagogy), have been a key battleground in education, from an almost dogmatic insistence that "things must be done in this way" to a open "inquiry-based" approach.  It surprises me that in all of these debates, which remain unresolved, little thought has gone into what actually constitutes a level.

Partly this may be because levels are seen as specific things which relate to a discipline. And yet, there are fundamental similarities between pedagogical approaches from learning Latin, music, or maths to astrophysics and medicine. There are stages, outcomes, assessments, and so on. One might think that these things are the products of the institutional structures around which we organise education. That might be true. But "levels" are nevertheless demonstrable irrespective of what assessment technique might be in operation, and their means of establishment have at the very least a family resemblance.

The most interesting and ancient of the levelled approaches is the "Gradus ad Parnassum". This refers to a range of different pedagogical approaches in different subjects. I became familiar with it through musical education, because it was the name of a treatise on counterpoint by Johann Fux. Fux's approach to counterpoint was to present learners with progressively complex exercises for them to complete. Because music is very abstract, these exercises are interesting because they present an almost paradigmatic case of the differences between one level and the next.

The basic idea is to write countermelodies to a given melody written in very long notes (called a Cantus Firmus). First, each long note is accompanied by one other long note which harmonises with it and whose construction must obey simple rules which form the foundation of the rules for the rest of the exercises. Secondly, each long note is accompanied by two shorter notes in half the rhythm. Then it is done by fours, and so on. Gradually the students learns the fundamental rules and how to mix combinations of shorter and longer notes over the original melody. The resulting music sounds like Palastrina. The technique was used by generations of composers who followed.

Fux's Gradus is interesting because each level has a certain completeness. The completeness of one level leads through the expansion and complexification of the technique to the next level. It would be very interesting to explore language pedagogies, and maths pedagogies for similar patterns of completeness. But I'm particularly interested in what this "completeness" at each level is.

It is not, I think, a construction of a particular accomplishment. That I think is an epiphenomenon. Somehow, by the performance of one level, a kind of 3-dimensional construction is made which eventually determines that the particular level is exhausted in possibilities. In other words, at a certain point, what happens next must be to stop at this stage, prepared to move on to the next stage. It may not be so different from a level in Space Invaders - and there there is a clue. What marks the end of a level, but the construction of Nothing. The invaders have gone, and so we begin again.

Open Source Canvas as an Educational Institution Innovation Platform

I’m spending a lot of time with Instructure’s Canvas at the moment. To be honest, I don’t much care for VLEs, and certainly have little interest in Instructure, who seem to be on a corporate path to datafying the university (although I strongly suspect this won’t work, so I relax). But Canvas itself – as software, interfaces, services, analytics, etc – is really interesting. My university has bought the top-of-the-range all-singing hosted version. But Canvas is open source, and you can download and install it from https://github.com/instructure/canvas-lms

It’s a bit fiddly to install, but it does work – all it requires is a Linux machine, and you follow the 

instructures… sorry, instructions 😉

It actually works very well. What you get is not just a VLE. You get a service-oriented framework for education, upon which the VLE interface sits. Theoretically, you could build your own interface.

But then look at what the services do: https://canvas.instructure.com/doc/api/

It’s really cool – I was able to automatically generate content, delete stuff, create accounts, generate users.. in fact, anything that can be done from the interface can be done programmatically.

Then there’s the LTI integration. New tools, new integrations, huge possibilities.

And then there’s the Graphiql query language for analytics.

This is very impressive. I’ve been trying to think what this is like.

It’s like a standardised platform for doing all the kind of administrative things that we want to do in education, but having a coherent and standard set of web-services for hooking in cool stuff behind the scenes. So run machine learning services in the background, or agent-based models, or new analytic tools which spit their results straight back to learners, or personalised learning which self-adapts to user engagement. But whatever you do, you can exploit the standard Canvas ways of communicating with students, including mobile notifications, apps, etc.

I think (although I’m not sure) Canvas feels like the first time we have had educational technology which is effectively a standard service-oriented platform that tunes in to the way educational institutions work. It’s like an Eclipse Rich Client Platform (remember that?) for educational institutions.

Am I getting carried away? I don’t know – but I want to find out!

From Radical Constructivism to Dominic Cummings: What's wrong with Cybernetics?

The pro-Brexit lobby at the heart of the UK government possess a powerful arsenal of conceptual and epistemological tools which have effectively been "weaponised" in ways which would have mortified their inventors. Dominic Cummings knows his systems theory (as a cursory glance at his "Thoughts on Education and Political Priorities": https://dominiccummings.files.wordpress.com/2013/11/20130825-some-thoughts-on-education-and-political-priorities-version-2-final.pdf). Cummings is not the first to turn cybernetics to bad ends - "philosopher" Nick Land has been writing pretty odious stuff for quite a few years, and it turns out that a big fan of Land is Andrew Sabisky who is coming under pressure for his somewhat insane views on eugenics (https://www.bbc.co.uk/news/uk-politics-51535367) - something for which Cummings also has a penchant. Hayek got there first with the dark side of cybernetics, of course, but this new breed is not as intelligent and more dangerous (Hayek was bad enough!)

It's all deeply troubling. Many of the inventors of these tools were German émigrés, horrified by Nazism, helping with the war effort by developing new weapons, and wishing for a better world. Wiener however knew that what they were doing was dangerous. His "The Human Use of Human Beings" reads like a prophecy today. Wiener's immediate fear was nuclear annihilation, but the likes of Cummings and his crowd are in his sights as the enslavers of humankind.

10 years ago I was at the American Society for Cybernetics conference in Troy, NY, which was attended by Ernst von Glasersfeld - one of the last remaining figures from cybernetics's early period, and an important thinker about education. Glasersfeld,  by then very old, gave a short address which summarised his philosophy; he died a few months later. You can read it here: http://www.asc-cybernetics.org/2010/?p=2700

It is such a clear exposition of cybernetic concepts that it invites a critical reflection: "is Cummings here?" - is there something in these ideas which opens the door to a fascism which would have mortified von Glasersfeld? I have to say, I think there is.

Von Glasersfeld made the clearest statement that cybernetics is fundamentally about constraint: as a science it is focused on "context". But as a science, it carried with it a clear conception of what is rational and what is metaphysical - and this is the main meat of the talk. Von Glasersfeld talks of the "pious fictions" of realists who insist on an external mind-independent reality. This, he states, cannot be science. Science, by contrast, exists in the rational process of coordinating understanding within constraints. As such, it cannot gain any kind of "objective knowledge".

But this sentence is the most interesting:

Only painters, poets, musicians and other artists like mystics and metaphysicians, may generate metaphors of reality, but to comprehend these metaphors you have to step out of the rational domain.

"Outside the rational domain"? What does that mean exactly? From what kind of context does von Glasersfeld make the judgement as to what is "rational" and what is not? This is framed by the existing institutional context of science and universities. Here we have embodied the problem of "two cultures" - and from there, we are on a slippery slope to Cummings.

Feelings are not rational. Social alienation is not rational. Experience itself is not rational. Yet some "rational" force allows us to make the distinction between what is and isn't rational, rejecting the irrational as a "pious fiction". This is how one can play the game of "Take back control" or "Get Brexit Done", treating feelings as if they are "rational" constructs of a communication system which is malleable to someone else's will.  It turns out that this is the pious fiction we should most fear. The artists, by contrast, speak the truth.

Where is the problem? It lies, I think, in a kind of two-dimensionality in the way that we think of communication. Cummings is quite keen on Shannon - at least insofar as it underpins data analysis.  But Shannon, lucky genius that he was, had a two-dimensional information transmission problem in front of him: a sends message to b over a noisy medium; b interprets and responds. But even in Shannon this isn't quite as two-dimensional as it seems: a and b are "transducers" with a "memory" (see Shannon and Weaver, "Mathematical theory of communication" (1951)). They were very pale representations of people. This meant that there was a limit to what could be communicated - and what could be constructed.

In Von Glasersfeld's world, the form of conversation occurred through the interaction of constraint produced through the communication of agents. Conversation and meaning emerged through a haze of "Brownian motion". It was almost arbitrary in its emergence, only recognised to be "meaningful" by us "observers". There are many problems with this view, the deepest of which is the assumption that within complex systems, the emergence of form and meaning is the result of an "arbitrary" process.

Years of attempting to simulate music from arbitrary processes have only produced bad music. It seems that the processes at work within the artist are no more arbitrary than the movements of electrons through a diffraction grating: there is an underlying pattern. But it's not the bands of the diffraction pattern that are interesting. It is the space between them: what is there? Nothing.

Appreciating this leads to a profound question about "constructivism" and indeed "radical constructivism": ok - so you can construct "stuff" in the world... but how might you construct "nothing"?

Without "nothing" there would be no pattern. Cummings, Land and co. know the deep magic. But the deeper magic is how to make nothing (channelling Narnia!)

To cut the story a bit shorter, "nothing" is mathematically realizable. William Rowan Hamilton's discovery of quaternions in 1843 was really the beginning of adventure into nothing which we have not absorbed yet. The quaternions are a 3-dimensional complex number which is anti-commutative. Hamilton's genius was to see that in order to represent the world in 3 dimensions, anti-commutativity was essential. But more importantly, the quaternion arithmetic allowed for expressions where a²  = 0. So 3-dimensionality and nothingness are fundamentally connected. But we knew this: ever heard of a "vanishing point"?

Von Glasersfeld had no way of constructing nothing. I asked him, a couple of years before when he gave a talk about learning in Vienna, that it was all very well to explain learning in the way that he was. But where did the drive to learn come from? He didn't really have an answer. Maybe he was tired. But I'm suspicious that he didn't want to think about it.

So Cummings and Land are exploiting a body of theory which is profoundly incomplete and two-dimensional. It's dangerous because it is two-dimensional and the real world isn't. The world of feelings, art, poetry and music are not some irrational boundary of a rational systems world. They are the third dimension in a world of natural information which cybernetics has not yet found a way to describe. It may become very important that we highlight these scientific shortcomings. 

Brains and Institutions: Why Institutions need to be more Brain-like

I was grateful to Oleg for pointing out the double meaning in Beer’s Brain of the Firm last week: it wasn’t so much that there was a brain that could be unmasked in the viable institution; firms – institutions, universities, corporations, societies – were brains. Like brains, they are adaptive. Like brains they do things with information which we cannot quite fathom – except that we consider our concepts of “information processing” which we have developed into computer science – as a possible function of brains. But brains and firms are not computers. That we have considered that they are is one of our great mistakes of the modern age. It was believing this that led to the horrors of the 20th century.

So what is the message of Brain of the Firm? It is that firms, brains, universities, societies share a common topology. In the Brain of the Firm, Beer got as close as he could to articulating that topology. It was not a template. It was not a plan. It was not a recipe for effective organisation. It was not a framework for discussion. It was a topology. It was an expression of the territory within which distinctions are formed. Topology is a kind of geometry of the mind.

Universities are particularly interesting examples. Because they are made of brains, and because their work is meant to be the work of their constituent brains. Universities present an example of where the “brain-organisation” sometimes goes right, but more often goes wrong. Why does it go wrong? Because we draw our distinctions in the wrong way – most often believing the institution to be the “organisation chart” – which is always a recipe for disaster.

Governments and states are alternative examples. One of the things we talked about was the general antipathy to the big state. The apparent failure of the Corbyn project is a hangover from the general disbelief in the big state. Johnson’s message is a throwback to Thatcher’s message – ironically this time marshalling the support of those who Thatcher hurt the most. But really, this message – the disbelief in the big state – never went away. After Stalin’s Russia, there appeared nowhere for the big state to go. But ironically, Stalin’s Russia was really a small state masquerading as a big one.

Today the world is full of would-be Stalins. Individuals wanting to impose their brains on everyone else, wanting to diminish the power of every collective unless it suits themselves. The model is repeated from corporation to university to city to state. But in the end, it will not work except to destroy its own environment (which it is doing very effectively), and it will not work because the distinctions are drawn in the wrong place.

The real question we should ask ourselves is this: How do brains work and how should organisations work to emulate them? Technology almost certainly gives a glimpse.

If there is a key feature of Beer’s fundamental topology it is the difference between the inside and the outside of a distinction. I wonder if in fact Spencer-Brown wasn’t influenced in his mathematics by Beer. I suspect Beer had the insight of Spencer-Brown’s most powerful idea first.

If you want to maintain any distinction then you must have a metasystem. Why? Because all distinctions are essentially uncertain, and there must be a mechanism – there must be the other side of the Mobius strip – to maintain the coherence of the distinction.

What must the metasystem do? Well, one of the things it must do is negotiate the distinction of the inside with the environment. In essence it has to determine what belongs inside and what belongs outside and to maintain this boundary. If it doesn’t do this, then the distinction collapses. So there must be a process of engaging, probing and modelling the environment.

This is System 4

The other thing the metasystem must do is to manage the internal operations of the system – its own internal distinctions. This is System 3.

Then it must balance the balancing operation of System 4 and System 3 This is System 5.
So very quickly we arrive at this topology.

But this is the topology of a distinction, and within any topology there are further distinctions. The point is it unfolds a fractal structure. This ultimately is the fractal structure of the inside which must be balanced with what is perceived as the fractal structure of the outside.

The challenge is to operationalise this.

For many who pursued the VSM, the operationalisation ended up as a kind of consultancy – a way of talking to organisations to give them a bit more internal awareness. I guess this was fine -  and it created a bit of work of cybernetics people. But ultimately this was empty wasn’t it?

How do we do better?

We need to come back to brains and firms. Within the brain, we have very little knowledge of what happens – particularly as to what happens with information. Obviously there are ECG monitors and stuff but they simply attenuate whatever complex activity is going on into graphs that show some kind of snapshot of the dance that’s really taking place.

We can see much more of “information” if we look at communications. Then what do we see? We see massive amounts of redundancy in our communication. We see pattern infusing everything, catching the attention of analysts and clairvoyants. The clairvoyants are usually more value because they tune-in to something deeper.

The deeper problem lies in the way we are able to analyse and examine the patterns of communication. It’s not as if we are short of data – although there can always be more. It is more that we do stupid things with the data we collect. Typically this involves attenuating out most of it and using an attenuated dataset as an exercise to make stupid decisions.

It’s rather like a brain with dementia. Important parts of the processes of maintaining information flows throughout the whole organisation are damaged – in the institution’s case, by technology – and consequently the selection mechanism for adaptation is impaired. Consequently the poor individual afflicted with this, steps into the road without bothering to check the approaching double-decker bus.
Our institutions are doing a similar thing for a similar reason. Key parts of their information processing apparatus are impaired which means that the selection mechanism for their future adaptation isn’t working.

So what is a selection mechanism for future adaptation? It is precisely what Beer, influenced no doubt by Robert Rosen, called an “anticipatory system”. It is the system’s model of itself. Now a model of oneself in time must be a fractal.  The only way the future can be predicted is through its pattern of events being seen to be similar to the past.
That is not that actual events repeat (although of course they do), but it is that the pattern of relations between particular events tend to repeat.

But we need to understand fractals. They are not really two-dimensional pictures. They are three-dimensional pictures. Long before we knew about fractals we knew the concept from the hologram – that encoding of time and space into a 2-dimensional frame where the self-similarity of the frame was its key feature.

The reason why fractals are so important is that our approaches to information and measurement are essenatially 2-dimensional. Look at Shannon’s diagram to see this. The deep problem with 2-dimensionality is that it has no concept of “nothing”. In Shannon, any symbol exists against the constant background of a not-symbol, but we have no way of expressing the not-symbol.
True nothingness means making things disappear. It turns out that the only way we can make things disappear is by working in three dimensions. The fractal is an encoding of three dimensions in which “nothing” is written through like a stick of rock. Nothing is what makes the pattern.
All human behaviour in institutions is really about nothing. Or rather, it is about the attempt to grasp nothing from something. In the way that a piece of music eventually selects its ending – and silence – all our behaviour seeks a kind of resolution. Every conversation seeks a resolution. Every interminable meeting frustrates because its ending frustrates.

But if we want to see nothing, then we have to work with its encoded representation – the fractal.
Our mathematical approach to information – to computer information – can provide a glimpse of nothing. Indeed, our approaches to machine learning, which are beginning to show behaviour that is rather like conscious behaviour, are providing a glimpse – they too are fractal.
If we want to see nothing, then we need an encoding strategy whereby data is represented in a way where nothing might be analysed and considered.

If we want viable institutions, then we need viable individuals. If we want viable individuals then we need a way of encoding the communicative behaviour of individuals in a fractal which can reveal the underlying selection mechanism for optimal future development. It would not be a surprise if the optimal selection mechanism for individual development involved communication with other individuals. And it would not be a surprise if the optimal collective development of a group of individuals entailed the preservation of information between them.

What do we have? Monasticism?

Why the current phase of Machine Learning will fail

Over the last two years I've been involved in a very interesting project combining educational technology assessment techniques with machine learning for medical diagnostics. At the centre of the project was the idea that human diagnostic expertise tends to be ordinal: experts make judgements about a particular case based on comparisons with judgements about other cases. If judgement is an ordinal process, then the deeper questions concern the communication infrastructure which supports these comparisons, and the ways in which the rich information of comparison is maintained within institutions such as diagnostic centres.

Then there was a technical question: can (and does) machine learning operate in an ordinal way? And more importantly, if machine learning does operate in an ordinal way, can it be used as a means of maintaining the information produced by the ordinal judgements of a group of experts such that the combined intelligence of human + machine can exceed that of both human-only and machine-only solutions.

The project isn't over yet, but it would not be surprising if the answer to this question is equivocal: yes and no. "Yes" because this approach to machine learning is the only approach which does not throw away information. The basic problem with all current approaches to machine learning is that ML models are developed with training sets and classifiers as a means of mapping those classifiers automatically to new data. So the complexity of any new data is reduced by the ML algorithm into a classification. That is basically a process of throwing away information - and it is a bad idea, which amplifies the general tendency of IT systems in organisations which have been doing this for years, and our institutions have suffered as a result.

However, not discarding information means that the amount of information to be processed increases exponentially (in fact it increases according to the number of combinations). It doesn't matter how powerful one's computers are, an algorithm with a computability growth rate like this is bad news. Give it 100 images, and it might take a day to train the machine learning for 4950 combinations. 200 gives 19500, 300 gives 44850, 400 gives 79800, 500 gives 124750. So if 4950 takes 24 hours, 500 will take 600 hours = 25 days. It won't take long before we are measuring the training time in months or years.

So this isn't realistic. And yet it's the right approach if we don't want to discard information. We don't yet know enough, and no amount of hacking with Tensorflow is going to sort it out.

The basic problem lies in the difference between human cognition and what the neural networks can do. The reason why we want to retrain the ML algorithm is that we want to be able to update its ordinal rankings so that they reflect the refinements of human experts. This really can only be done by retraining the whole thing with the expanded training set. If we don't retrain the whole thing, then there is a risk that a small correction in one part of the ML algorithm has undesirable consequences elsewhere.

Now humans are not like this. We can update our ordinal rankings of things very easily, and we don't suddenly become "stupid" when we do. How do we do it? And if we can understand how we do it, can that help us understand how to get the machine to do it?

I think we may have a few clues as to how we do this, and yes I think at some point in the future it will be possible to get to the next stage of AI where the machine can be retrained like this. But we are a long way off.

The key lies in the ways that the ML structures its data through its recursive processes. Although we don't have direct knowledge of exactly how all the variables and classifiers are stored within the ML layers, we get a hint of it when the ML algorithm is "reversed" to produce images which align with the ML classifiers, such as we see with the Google Deep Dream images.

These are basically fractal images which reflect the way that the ML convolutional neural network algorithm operates. Looking at an image like this:

we can get some indication of the fractal nature of the structures within the machine learning itself.

I strongly suspect that not only our consciousness, but the universe itself, has a similar structure. I am not alone in this view: David Bohm, Karl Pribram and many others have held to a similar view. Within quantum mechanics today, the idea that the universe is some kind of "hologram" is quite common, and indeed, the hologram is basically another way of describing a fractal (indeed, we had holograms long before we could generate fractal images on the computer).

What's important about fractals is that they are anticipatory. This really lies at the heart of how ML works: it is able to anticipate the likely category of data it hasn't seen before (unlike a database, which can only reveal the categories of data it has been told about).

What makes fractals awkward - and why the current state of machine learning will fail - is that in order to change the understanding of the machine, the fractal has to be changed. But in order to change a fractal, you don't just have to change one value in one place; you have to change the entire pattern in a way in which it remains consistent but transformed in a way where the new knowledge is absorbed.

We know, ultimately, this is possible. Brains - of all kinds - do it. Indeed, all viable systems do it. 

Brexit Lions and Unicorns

Orwell's essay "The Lion and the Unicorn: Socialism and the English Genius" reads today as very old-fashioned and jingoistic. And yet, like all great artists, Orwell accesses something of the nature of life  - both at the time he was writing ("As I write, highly civilised human beings are flying overhead, trying to kill me.") which we are seeing reflected back at us in a rather unedifying way with Brexit.

Much that he identified in 1941 is still true.

"There is no question about the inequality of wealth in England. It is grosser than in any European country, and you have only to look down the nearest street to see it. Economically, England is certainly two nations, if not three or four."

Except that of course this inequality has been exported to many other countries. But what about "patriotism" which his essay is really about? What does that mean?

It seems that patriotism is very problematic and confusing. It is uncomfortably close to the "nationalism" or "populism" (what does that mean?) that we see among the Brexiteers or the Trump supporters. Orwell's point is to say that despite the differences between the rich and the poor, or the different nations of the UK (I would like to think he would apply this across our multicultural society today) a country can be united if it feels its "home" to be under threat. I doubt this is a "national" instinct but a human one - we see it in extraordinary human acts of courage and compassion in the wake of terrorist atrocities (for example) the world over. Can we explain it? No - but to "explain" it as a "national characteristic" is both tempting and facile - that's where jingoism comes from.

Where does the "threat" which mobilises everyone come from? Orwell is very clear that it is not among the individual "highly civilised human beings" who are trying to kill him.

"Most of them, I have no doubt, are kind-hearted law-abiding men who would never dream of committing murder in private life. On the other hand, if one of them succeeds in blowing me to pieces with a well-placed bomb, he will never sleep any the worse for it. He is serving his country, which has the power to absolve him from evil."

I imagine whether one might say this about some members of ISIS. So much depends on what we consider to be "civilised".

But it's not fanatics who scare us: "What English people of nearly all classes loathe from the bottom of their hearts is the swaggering officer type, the jingle of spurs and the crash of boots.". It's the aristocracy who, in Europe adopted the goose-step as their "ritual dance" - and his point that it is hard to imagine a Hitler in the UK relies on the fact that "Beyond a certain point, military display is only possible in countries where the common people dare not laugh at the army." Orwell argued that we needed socialism to counter the aristocratic swagger which leads to people like Hitler. What he called elsewhere the "arrogance of superiority".

Today in Europe we don't see military officials swaggering in uniform, although the police often now carry guns. But there is still the swagger of authority everywhere - and I think this really lies at the heart of the Tory Brexit which has just taken place. Orwell may be right in what unites rich and poor being authoritarianism - it absolutely fits the rhetoric of Nigel Farage, Dominic Cummings and others. But there is no goose-stepping, and all that the Brexiteers complain of is "Brussels bureaucrats". That's a funny kind of goose-step!

But it may be one nonetheless. This is goose-step in the techno age. It is the goose-step of free-market capitalism, technologically-driven oppression, surveillance and artificially-imposed austerity. We are all marching to its beat, and most of us - the losers - hate it.

We know, deep down, that the likes of Johnson, Trump and Farage are really the commanders of this robot dance which so many detest. And we know that they know that declaring hatred for it on the one hand, and ramping-it up on the other is the game. But we are caught - because you cannot call them out without denying that the uniting hatred of authority is fundamentally true. It's a massive double-bind.

So how do we get out of this mess? Orwell's answer was socialism - and in many respects he knew that the founding of the NHS and welfare state was inevitable after the war. It's taken over 70 years to re-impose the goose-step in the techno age in a far more complex and uncertain form where it drives everything, distorting that early socialist ideal to move to its beat.

One of the striking things about the EU and Brexit, Westminster and the election, is that everyone has taken it so seriously. Perhaps we shouldn't take any of this seriously. Perhaps the game is to get us to take seriously things which aren't at all serious. It's like the population who is afraid to laugh at its army. If we don't take our parliaments - whether national or international - seriously, what are they? They are - it all is - irrelevant. Now look at how hard the media - the press, the national broadcasters, the social media companies - are trying to convince us that this isn't irrelevant!

That's the mark of swaggering authority - to persuade the people that what is irrelevant is the most important thing in the world. It's a con.