Revolutions Of Scientific Structure, The

Colin G Hales

• Publisher: WSPC
• Year: 2014
• eBook ISBN: 9789814616263
• Topic: Biological Sciences
• Subtopic: Science General
• Index: Biological Sciences

This book discusses two main cultural problems behind the failure of machine consciousness and artificial general intelligence (AGI) projects over many decades.

The first problem recognizes that building a conscious AGI means building an artificial scientist. The book identifies the responsible pitfalls in mainstream scientific behavior and eliminates them by proposing a new operational framework for scientists called “Dual Aspect Science”.

The second problem arises because scholars involved in machine consciousness and AGI essentially aim to replicate brains with computers. They are demonstrably not doing this, and this failure has been prevalent since the rise of computers. Instead, the book discusses the possibility of doing real empirical neuroscience by means of artificial materials that literally do what the brain does.

Inspired by Thomas Kuhn, one of the most influential philosophers of science of the twentieth century, this compendium proposes a fresh perspective on machine consciousness, on AGI and, more generally, on how the machinery of science might need to change to accommodate it.

Contents:

• Preamble
• Introduction
• Consciousness
• The Route to Normal Science
• ‘Normal’ Science
• The Great Blockage
• Cultural Learning Theory for Scientists
• The ‘Law of Scientific Behaviour’
• The Biology of Belief: Statement Formation
• Hierarchy, Emergence and Causality
• Dual Aspect Science
• Scientifically Testing for Consciousness
• The Kuhnian Take: Wrapping Up
• Machine Consciousness and DAS

Readership: Researchers, academics, professionals and graduate students in artificial intelligence, robotics, neuroscience, computational physics and nonlinear science.

Chapter 1

Preamble

We live on an island surrounded by a sea of ignorance. As our island of knowledge grows, so does the shore of our ignorance.

John Archibald Wheeler [Horgan, 1992]

I am an engineer and, latterly, a scientist and I like to think of myself as living on Wheeler’s beach. Here on Wheeler’s beach with my scientist colleagues, surfing the tides of ignorance, one way to sort out scientific from unscientific discourse is to see if a proposition, say proposition-X about the natural world, comes somehow inextricably laced with history and personal context. The more there is, the less credible is proposition-X.

To a scientist this makes perfect sense and works quickly. It’s part of the regime of objectivity. If proposition-X is to be valid science, then it is either going to be predictive of the natural world or it isn’t. Observational evidence is decisive, and no amount of history or background of the proposition-X enthusiast is going to change that. If someone is to convince you of proposition-X and somehow its validity is dependent on how it originated, how it is communicated (including the language used) or who communicated it, then X may be quite valid in some context, but it’s not a scientific proposition and it cannot have arisen from scientific behaviour. The skepticism meter ‘red-lines’, and proposition-X is set aside for a different world and another time.

These are the thoughts that I must overcome to populate this preamble with personal background, and I confess the process does not sit well with my training as a scientist. Nevertheless, in this instance I must persist. In the upcoming chapters I propose a rather far-reaching alteration to scientific behaviour. The personal history speaks to motivation and the motivation led me directly to a realm of inadequacy and anomaly in science that, it is to be argued, necessitates the proposed change to science. The inadequacy and anomaly were encountered when I tried to do a certain kind of science. It is what happened to me and would happen to anyone that tries to walk the path involved in solving the problem I am trying to solve. My journey has become scientific evidence about science itself, not scientific evidence supporting one of the day-to-day outcomes produced by science. If I leave out the history then the motivation and apparent necessity for change to science would be diminished. To impoverish the discourse in that way, even accidentally, would do a disservice to those involved.

I love to build things. I want to build machines that exhibit the same robust, adaptive learning behaviour as biology including, ultimately, human-level capacities. I come from an engineering background involving making computer-controlled machines operate usefully in the natural world of business/industry. I had an entire career in it. In their pitiful adaptability and their fragility in the face of novelty, the machines I automated were woefully inadequate. Indeed I ran a business based on those inadequacies. Business problems necessitated the use of highly trained experts to cajole machines into usefulness and then keep them there. There was money to be made in this. However, with market pressures, the pace of change made their inadequacies ever more apparent. Throughout it all I remained increasingly aware of the amazing difference, in terms of handling novelty, between our most sophisticated machines and the lowest members of the animal world with a nervous system. Then, as circumstances permitted, I underwent a bifurcation and took a trajectory towards solving this problem in an academic setting. This is the way I describe the process, and the reasons I use this terminology will become apparent later.

In effect, I have become a scientist to solve one and only one problem; to answer the question “How is it that biology does so effortlessly what our machines consistently fail to do – handle novelty ?” Once answered, I can help build better machines.

It turns out that my chosen problem is generally officially recognised to be about 60 years old, although the idea of it has a history evident from the ancient Greeks onwards. In 1956 at the almost legendary ‘Dartmouth Conference’, the problem of intelligent machines was christened as the problem of artificial intelligence (AI) [Moor, 2006]. AI headed towards machines that might potentially have human-level intelligence or better. We all know this program of work has, to date, failed. In its stead, what has been its successes are now better called narrow-AI. A computer can beat a human at chess. A computer can beat a human at Jeopardy. Sixty years of faddistic fashion-centric computing has given us this much. Powerful narrow-AI in a range of guises now sits invisibly at the core of our lives. This pattern of AI progress happened against a background of failure by proponents of solutions to the original goal of human-level AI. There has even been the odd ‘AI Winter’. Burned out AI bandwagons litter the science landscape [Anderson, 2005; Beal and Winston, 2009; Brooks, 2008; Gelertner, 2007; Holmes, 2003; Hopgood, 2003; Leake, 2006; Mullins, 2005; Reddy, 1996; Shi and Zheng, 2006].

The declining investment in those claiming access to the goal of the original project – human level AI – created a group of marginalised and impoverished enthusiasts. In the last 10–15 years or so, however, there has been a resurgence that is probably due to the almost blinding escalation in computer power resulting from a corporate policy called Moore’s Law. Sheer computational grunt has people buzzing with expectation. This resurgence might also have something to do with the fact that the older generations with memories of AI failure are exiting, and a new generation, unencumbered by the pain of a half century of failure, sees more hope in what has become a rebadged area: ‘Artificial General Intelligence’ or AGI. AGI now has at least one journal and a conference and has become an active community. It still produces narrow-AI outcomes, however, not AGI (yet, anyway).

My mission, then, according to the current jargon, is to create practical AGI. I have my ideas, and I am doing the science. As this is being written, a supercomputer or two is choking on my simulations, and with my physics hat on I am building prototypes for a new kind of neuromorphic chip technology. None of which has a direct bearing on what is going on here. I said I am here to solve the problem of AGI. What this book is mostly about, however, is not AGI. It is about what happens when science tries to cope with a real solution to AGI.

Human-level AGI means that in principle, the AGI could do anything a human can do. One of the things a human can do is science. Therefore, to build an AGI is, amongst many other possibilities, to build an artificial scientist. One of the essential intrinsic qualities of scientific behaviour is that it is science’s job to confront novelty – the unknown. Not only that, it’s testable because scientists provide evidence of contact with novelty by delivering it in the form of a novel ‘law of nature’. It seems perfect as a target model for my AGI project. As an engineer with a useful behavioural benchmark of scientific behaviour as my goal, I know I can proceed by borrowing from a ‘science of scientific behaviour’, internalise its ‘laws of nature’ and then build it. Simple.

The problem I encounter is this: The first thing you discover is that these ‘laws of the natural world of scientific behaviour’ have not been written down by scientists. They are not in text books. Scientists are unaware of it. There are no international organisations designed to create or manage such a thing. There is no science discipline charged with responsibility for it. Scientists acquire their craft by imitation of mentors. Having recently done a PhD I have just experienced this very process. I am and have first-hand evidence. The next important thing to note about scientists is an ability to observe (perceive the natural world with those aspects of mental life we call perceptual fields such as vision). My AGI must become a scientific observer, amongst other things. Therefore I need a science of the scientific observer. If you investigate science you will find, throughout the entire history of science, that the observer is systematically excluded from all scientific accounts of nature. Observations are used to account for what is observed. Observations do not explain or even predict an observer. Indeed for the entire history of science the scientific observer has been presupposed.

This is the knot that I must undo in this book.

But the knot is even more interesting than this. Only in the last 20 years has science accepted that a scientific account of scientific observation (an ability to observe) can be standard, funded science of the familiar kind. This follows hundreds of years of exile as a spurned pariah. This nascent science happens in neuroscience. One of the names it travels by is ‘The Neural Correlates of Consciousness’. We scientifically observe using our consciousness. Consciousness is an identity with a collection of perceptual fields. They are our mental sensations. Take all sensation away and there is no consciousness. To supply a scientific account of scientific observation (an ability to observe at all, not the contents of an observation) is to supply a scientific account of consciousness. But this connection of two characterisations of the one thing as being both ‘consciousness’ and ‘scientific observation’, masks the bald fact that the one thing that science has eschewed for 2000+ years is the one thing that has been taken up and studied by science in a way that masks the reality of the connection. This, it is to be argued, is a fundamental anomaly in science that places science at the cusp of a necessary adjustment. We cannot have it both ways. Science cannot implicitly study the scientific observer ‘scientifically’ on one hand (calling it a science of consciousness), and then methodologically, by omission, eschew it as an explicit account of a scientific observer while presupposing the observer to do it. This fundamental logical inconsistency is right there for everyone to see. There are also a range of other associated side-effects to be discussed later.

This book is called ‘The Revolutions of Scientific Structure’. The name is a play on words of the title of a great book on revolutions in science. I need to explain why I have chosen the title. In the context of the title, the word ‘revolutions’ is only just plural. It is to be shown that in this exquisitely specialised area of the science of scientific behaviour, there can be only two revolutions. Whenever it started (probably in the ancient Greeks that hit upon the idea that the natural world can only be understood by observation, not through authority), the first revolution put the observer at the centre of scientific behaviour and simultaneously stripped the observer from all the resultant descriptions of nature. These descriptions we regard as the cumulative output of this specialised human behaviour, scientific behaviour. But these ‘descriptions’, sometimes called ‘laws of nature’, are merely descriptions. The process presupposes an observer that describes the natural world’s appearances. So far, so good. One revolution so far, one kind of scientific behaviour so far.

But things have now changed. I will argue that as a result of twenty years of the science of consciousness, we have now (albeit tacitly) targeted the scientific observer as a problem to be solved by a science that systematically presupposes an observer. But the tools of the first revolution are all we have. We are all logically compromised at a fundamental level. Science has been presupposing the observer for so long it has no explicit mechanism for navigation of the change needed to bring the observer into the scope of a more mature science that can also account for an ability to scientifically observe. At the time of writing we are all within this anomalous cusp in the history of science and are essentially unaware of it. We need to become aware of it. It’s going to happen. We cannot avoid it. We either need to account for the observer with what we do (which is indefensible question-begging that explains nothing), or change what we do to allow a context of accounting for nature in which it makes sense to explain the observer/consciousness.

The reason it is possible to scientifically detect this state of affairs is due to the work of Thomas Kuhn who, in 1962, published an examination of hundreds of years of science history to show us the specific signs and dynamics of scientific revolution. In ‘The Structure of Scientific Revolutions’, Kuhn delivered the hallmarks of scientific paradigm change [Kuhn and Hacking, 2012]. Science progresses, in the more modern parlance, via the trajectory of equilibrium (called ‘normal science’) punctuated by radical changes in trajectory called bifurcations, which occur at instants of high instability (in science this has the appearance of intractable anomaly) [Gould and Eldredge, 1972]. Long periods of incremental normal science encounter anomaly and this triggers a crisis. Subsequently there emerges a qualitatively and quantitatively different form of science in which the offending anomaly is gone. From the vantage point of the new science, the old science is not seamlessly commensurable with the new (it requires a deal of work that is unfamiliar to the old system), and the old science does not and cannot predict the new. The entire process, Kuhn correctly judged, is mediated by the mental phenomena of the scientists involved. This is a neurological phenomenon.

Here, with a view to facilitating the changes needed to account for the scientific observer, I have simply observed science itself, and applied Kuhn’s ideas to scientific behaviour, rather than the products of scientific behaviour. Not only have I found a way ahead, but I see that all the hallmarks of paradigm shift exist right now. All the signs are there. The revolution I predict is the second and fundamentally the last, of two. Once a scientific account of the scientific observer (consciousness) exists, whatever the world looks like after that is one in which the mental life of humans is predicted and explained, and is a normal target of problem solving. In that world it then becomes a capacity that can be deployed into machines if we wish, because we have the appropriate science to give to engineers to build. That science currently does not exist.

Note that this book does not attempt an explanation of consciousness. Rather, what is covered is what scientific behaviour must be in order that a real explanation of consciousness is accessible at all. What is delivered is a new framework for science, along with a new kind of scientific behaviour, not an explanation of consciousness. In a later chapter I will detail the new science framework along with how the new and the old interrelate. I will also be demonstrating that some scientists have been doing the new form of science for years without knowing it. The final structure for science is not ‘one or the other’. Science moves on with both the old and the new operating side by side, tied at the hip in the explanation of scientists.

The new framework is called ‘dual-aspect science’, thereby contrasting itself with what we do at the moment: ‘single-aspect science’. This is how and why there are specifically and only two revolutions in the structure of science. With two, scientific knowledge acquisition (scientific behaviour) is self-consistent, complete and both descriptive (what/predictive of appearances) and explanatory (why/reveals underlying causality). This contrasts with the past, where it has been inconsistent, incomplete and only ever described (what), never explained (why). In the new structure for science, I am proposing a new kind of ‘law of nature’. I am claiming that the existing kind must fail to account for the scientific observer, and because we continue to act as if it can, we scientists are logically compromised at the deepest level and are unaware of it.

The next advice from Kuhn serves as warning to those with whom my story may clash, and from whom I expect resistance. In this particular case it is the whole of science. This is no isolated detail within a specialised sub-domain of science. Bringing the observer into science affects every scientist, experimentalist, theoretician and the laity. If the proposition for a new science framework is logically sound and shows promise of being empirically proved effective, then Kuhn advises that a cohort of those willing to entertain the new paradigm will establish a camp on Wheeler’s beach, and that the rest of the resistant community will be found, over time, to be increasingly less able to defend their position. In the end, Kuhn advises, their defence of the old way will cease to be regarded as scientific. Paraphrasing Max Planck, sadly sometimes we must admit that as one quintessentially human endeavour amongst many, “science advances one funeral at a time”. Reason can be a casualty in the war of the shifting of paradigms.

Finally, for me, the nice thing about this is that I don‘t need anyone to agree with me to get on with my AGI work. I can get on with it, and it will look like ‘normal’ science. The journal articles will speak in normal tongues and claims will be empirically justified as usual. What will be different is that in my case, the AGI design is informed merely by considering the obvious logical consequences of the new framework for science detailed in the following chapters. It gave me hints as to what the natural world might look like when it is delivering sensations to the subject (it looks like particular bits of a brain electromagnetically dancing in a particular way). My AGI approach simply replicates it inorganically. I am building inorganic brain tissue and it does not involve computers at all, like us - the natural general intelligence. My mission here is simply to deposit the reality of the modifications to science for others to sort out. I must get this out of my system. While others digest it (or ignore it!), I will simply be getting on with my AGI science.

If you read Kuhn’s book you will constantly find yourself mentally grating with an old-school gender bias that exceptionlessly associates male pronouns with the entire history of science. He, his, him, man, men. He may have meant it in a genderless manner. On the other hand, Kuhn was brutally thrusting a mirror in the face of scientists; a mirror of the historical record. In exactly the same way that so many other darlings were sacrificed in that mirror by the reality of the track record, the fact is that in the deep history of science paradigms, the actual coal-face of activity was populated mostly by men. If you read his words, however, you‘d think that not one woman ever did anything scientific. Perhaps Marie Curie might have rated a mention? I don’t understand the actual message behind this gender issue. Maybe it’s just a pair of sensitised 21^st century eyes viewing words of a different era. Those that knew him better may have more insight into this.

Once I was able to set the gender issue aside, Kuhn’s words became a kind of poetry to me. There are things expressed in Shakespearean ways that seem beyond the reach of improvement, or that to think I could improve would be some kind of art crime. There’s a kind of beauty in the way they touch the subject so deftly. So I decided, in the end, in tribute to his book, to let his actual words form a constant backdrop, guiding me through the pages. Perhaps the words will help me in my quest to persuade. Perhaps not. In any event, I hope you will enjoy them as I have.

You never change things by fighting the existing reality. To change something, build a new model that makes the existing model obsolete.

Buckminster Fuller

1.1Reading this book: The quick way

A summary is provided at the end of each of the technically detailed chapters. When combined with the first couple and last couple of chapters, the reader can access the essential messages of this book. For practical implementation of the dual aspect framework, see Chapter 11. For practical testing for consciousness, see Chapter 12.

Chapter 2

Introduction

In 1962 Thomas Kuhn published a landmark examination of millennia of science history to show us a stasis/revolution pattern in scientific progress [Kuhn and Hacking, 2012]. This came as a surprise to the scientists of the time, who were happily operating inside what they thought a cumulative, well oiled, trouble-free novelty-seeking machine. In Kuhn’s ‘The Structure of Scientific Revolutions’, he starts with:

History, if viewed as a repository for more than anecdote or chronology, could produce a decisive transformation in the image of science by which we are now possessed.

[Kuhn and Hacking, 2012, Page 1]

In this dense style he then goes on to do just that with the huge trail that is the historical record of science’s output, good and bad, lasting and short-lived, right and wrong, universal and specific. He looked at the record of the journey of the people that produced the science, their communities, what appeared to motivate them and how the machinery of science actually worked. Anyone looking at this kind of literature will discover that the scientific production line involves social/political circumstances, competition, back-scratching, collaboration, life goals, authorities, prestige, entrepreneurship, tradition, ritual, fashions, eminence, mentoring, preferences, prejudices, favourites, secrecy, branding, peer association … science has it all. Underneath all of this, however, Kuhn revealed science, throughout history, as having operational methods and commitments very similar, in principle, to those used today.

The acquisition of scientific knowledge, Kuhn demonstrated, proceeds through a process exactly like biological evolution by natural selection. In evolutionary terms, we’d say that during environmental stability, evolution preferentially upgrades animals to take advantage of environmental subtleties. These are the paradigms Kuhn calls ‘normal science’. Many small changes. Some would call it incremental science. I would call it a power-law graded equilibrium. In ...

Table of contents

• Cover
• Half Title
• Series on Machine Consciousness
• Title Page
• Copyright
• Dedication
• Preface
• Contents
• List of Tables
• List of Figures
• Chapter 1
• Chapter 2
• Chapter 3
• Chapter 4
• Chapter 5
• Chapter 6
• Chapter 7
• Chapter 8
• Chapter 9
• Chapter 10
• Chapter 11
• Chapter 12
• Chapter 13
• Chapter 14
• Bibliography
• Index