Customer Reviews

How to Build a Mind: Toward Machines with Imagination (Maps of the Mind)

5 star:		(3)
4 star:		(0)
3 star:		(2)
2 star:		(3)
1 star:		(0)

 

 

This book treats consciousness with a disturbing kind of ease. While Mr. Aleksander brings out the issues surrounding consciousness (if one can even do that!). All too often the issues are either one sided to too simplified to the point it makes the author look like he has his mind made up before he asks the question.

Another serious issue I have is the ratio of philosophers to text used. I haven't seen this blatant use of names since I read Bart Kosko's book on fuzzy logic. While Aleksander tries to model a "mind" on silicon he eludes defining consciousness while raising the ability of machines. From his book; "The key difference between the machine and the person is that the machine would be conscious of being a machine, whereas the person is conscious of being a living human." How are we even to guess when a machine is conscious of being a machine? Does my toaster "believe" it's a toaster?

I am not a philosopher but an engineer and I've studied neural networks and I do agree with his suggestion that emergent properties can arise from complex systems. While others see consciousness an emergent property of a neural network - I have yet to see evidence of this or... even an indication of this. If you haven't had any exposure to neural networks or philosophy AND you want to see a snapshot of the controversy surrounding the issues of consciousness THEN you might want to read this book.

The author's goal is to answer the question as to whether a (non-human) machine can imagine. Clearly he believes that machines can, and throughout the book he gives his reasons for believing so. Early on, he emphasizes to the reader that he is an engineer, but given the view by most that engineering is a practical profession, he also wants the reader to know that it is philosophy that permits a true understanding of the nature of machine intelligence and forms the proper context for addressing questions regarding the ability of machines to have an imagination. Indeed, research into machine imagination is considered to be a combination of engineering and philosophy. Those readers, including this reviewer, who find philosophical speculation a distraction to the actual construction of intelligent machines might not want to read further. However, there is enough discussion on the history of the author's involvement in the development of intelligent machines to make the book worthwhile to read. This is especially true for the author's discussion on the MAGNUS machine, which he considers to be a machine "driven by inner states." In addition, the author is very aware of the pitfalls of philosophical musings on the nature of consciousness and machine intelligence. One of these concerns the conflict between the use of mathematics and physics to promote true understanding, versus the insistence that such understanding can only be reached from the use of thought experiments and argumentation. Another problem, says the author, is the predilection of philosophers to deny or negate the thoughts of their predecessors, which stymies progress to true understanding and is to be contrasted with the more effective approach in scientific circles, where consensus can be reached based on available evidence. Lastly, the author believes, the drive to understand consciousness has driven philosophers to the embrace of mysticism, with a consequent rejection of quantitative approaches.

The design of non-biological machines with imagination is not only driven by curiosity, but also by the desire to shed light on the nature of consciousness itself, says the author. The actual implementation of conscious imagination in non-biological machines can assist in the understanding of how it is done in biological machines, or at least how they are to be contrasted. The mechanisms giving rise to imaginative consciousness may have common elements in biological and non-biological machines. The author wants to find what aspects of "artificial" imagination are in fact true for "real" imagination.

At various places in the book, the author includes hypothetical discussions and debates with various philosophers and notable persons in history. These are interesting for sure, but they distract the reader from the discussion on the actual engineering of conscious and imaginative machines. Philosophers who find machine consciousness an elusive or impossible goal will never be convinced by any arguments supporting this goal. It would be better if researchers in machine intelligence would declare a moratorium on philosophical debate and speculation, and instead get busy with the real goal of designing and constructing intelligent machines.

The author characterizes consciousness in a machine as being the ability to know where it is situated, as being an understanding of its origins, and having its own motivations for the making of decisions. These criteria don't really that seem to difficult to implement in non-biological machines, and as one reads the book it becomes more apparent with each passing page that the author does not consider the implementation of non-biological machine consciousness as being a problem of overwhelming difficulty. His optimism in this regard is very characteristic of those who work in the field of machine intelligence. Their efforts are admirable, and even though the engineering of consciousness in a non-biological machine may remain elusive in years to come, there is no doubt that various types of machine intelligence have been realized in some of the machines of today. We can only expect further advances, and the rise of new types of intelligent machines. Whether these machines meet our expectations is another matter, but they have already exceeded expectations in many cases. Conscious or not, the machines of the future will certainly be fascinating entities with which to interact.

Read this book only if you need to be convinced that machines will be capable of thought and imagination. It is quite philosophical and argues points that seem obvious to me. The book seems to be intended for doubters. The first 11 chapters get three stars from me.

On the other hand, if you want to understand how the brain works, turn to the last chapter (Ch# 12 On Being Conscious) it is an excellent summary of our current knowledge. This last chapter gets five stars.

"How to Build a Mind" by Igor Aleksander, does everything but tell you how to build a mind as its entanglements with philosophy leaves the reader questioning whether this book should indeed be categorized under neuroscience or under the philosophy of science. Succinctly stated it is a short almanac on varying temporal perspectives on the philosophical definition of the mind with nominal science thrown in every few chapters.

At the onset, I was enthralled at the prospect of being able to read the writing of famous neural engineer: Igor Aleksander. As a member of the Royal academy of Engineering who revolutionized the field of neural networks, Dr. Aleksander is known throughout the world as a leader in his field. As such I approached the book with the trepidation that stems from reading the thoughts of a man whose intellect far supersedes my basal one. Considering that Dr. Aleksander has invented the first `conscious machine "MAGNUS", the title of the book lead me to believe that this would be an expose' on that phenomenal achievement (as the title does read: `How to Build a Mind'). From the onset, the author spends a majority of the book in pointless and completely imaginary banter with the philosophers of yonder lore to no effect. Considering that these conversations are completely imaginary with no basis in history, it is difficult to fathom the literary value of such text that does nothing but add to volume of the book. As I read through countless pages of painful conjecture, I clung desperately to the possibility that there might still be some part of the book that spent time explaining the complexities that are associated with generating neural nets and the science that undergirds it all. Needless to say I was deeply disappointed as the science (both biological and computational) presented did not exceed the level of an average high school student which essentially prevented any meaningful exploration into the phenomenal achievement that is the development of conscious neural nets. Over 75% of this book is fundamentally a thought experiment into philosophically examining the definition of consciousness through the eyes of all the great philosophers of yester years.

The book is divided into twelve chapters which temporally progresses through either philosophy or the research career of Dr. Aleksander with chapters in which both are intertwined. To be fair to the author, Dr. Aleksander does outline his book appropriately in the first chapter with the initial disclaimer that the book was not written purely for the scientist as such an audience would be limited indeed. The second chapter is set in the Greek town of Miletus which was the home to the fathers of philosophy Anaximes, Thales and Anaximander. It is in this chapter that the author engages in a completely fictional and non historical account of a discourse between Thales and Anaximander and later between Anaximes and himself, each debating about the divinity of the soul, life itself and its relationship to objects. Following this fictional account, in the third chapter of the book the author shifts gears by covering the history of `cybernetics' from its inception in 1958 with multiple references to Norbert Weiner's pioneering work chronicled in the book "Cybernetics". Before dipping too much into reality, the author finds an incessant need to jump back into the foray of fictional discourses in the fourth chapter of the book titled "The Ghost of Aristotle". This chapter recounts Aristotle's trial in Greece with the accusations against Aristotle ranging from his influence on Alexander the Great to the charge of `impiety'. In this chapter, the author takes pot shots at religion with multiple references to theistic failures during Aristotle's time. In what is quickly becoming the classical structure of the book, the next chapter focuses purely on Dr. Aleksander's initial work in the field which started with his first research position at Queen Mary University. It is in this chapter that some semblance of neuroscience is presented with a concise summary of certain aspects of computational modeling and the biological underpinnings of neuronal function. The chapter following this adheres to the pattern that any reader will have come to expect: another fictional account. This time around, in the chapter titled "Liberating Philosophy", a whirlwind of make belief is experienced as the author interviews the likes of Rene Decartes, John Locke and David Hume about the perception and definition of consciousness and whether it can exist in an object. After this little imaginary dip into philosophy, the author then predictably returns to reality in the next chapter as he describes the first neural circuit model that he built called Minerva. The basic workings of Minerva are explained without any mechanistic details by simply stating that RAM's with three inputs and outputs were used to develop the system which had a limited ability to recognize letters represented by bulbs on a circuit board. The next chapter elopes into the realm of fiction again as the author imagines being in conversation with Dr. Wittgenstien (a strong proponent that machines cannot have consciousness) on the very notion that machines can indeed have a consciousness. From this fictional realm, the reader is bought back to reality as the author superficially touches on the development of his next neural net christened WISARD which was designed to overcome the failures of Minerva through the employment of a larger RAM backbone (256 Megabytes which at the time required a long time to even obtain from the manufacturers). WISARD as the author notes was still vulnerable to Alan Lightman's theory of combinatorial explosion. Following the lack of computational details about WISARD, in the following chapter the author provides a transcript of his discussions on the BBC show known as `Start the Week' about consciousness. This discussion was between the author and multiple scientific personalities well known in research circles, Susan Greenfield, Roger Penrose, Margaret Boden, Aaron Sloman, Francis Crick and Daniel Dennett. The chapter portrays the author as the lone star general who stood by the idea that machines could indeed be conscious. Following this proclamation by the author, the following chapter focuses on the development of MAGNUS which overcame all the challenges faced by WISARD. This machine was largely successful due to the implementation of three states, input, output and a middle state which remembered previous inputs. Such a design thus gave the machine an internal state which the author claims is representative of consciousness. The final chapter of the book examines the theoretical constructs from which consciousness is both understood and explained. These constructs the author purports is in line with the idea that machines can be conscious.

In looking back at the book, the few memorable lines are the recounting of Descartes famous statement, "Cognito ergo sum" (I think therefore I am) and Betrand Russels quote about Aristotle in chapter 4 of the book which stated " After his death, it was two thousand years before the world produced any philosopher who could be regarded as approximately his equal". While the author's prose is gifted indeed this book is anything but prose and as such, the lack of computational, mechanistic and biological details combined with fictional and imaginary accounts makes this book a challenge to read. Furthermore, the author digresses on numerous occasions, often through unashamed admission with further attacks the credibility of the message being narrated. The one aspect that the author does justice to is his interest and passion in philosophy that is reflected in the numerous quotes that prelude every chapter.

All in all, for anyone interested in a philosophical discussion about consciousness and its ramifications in the computational world should read this book with little expectation for any serious scientific explanation of the development of any of the neural nets that the author has received notoriety for.

There's an interesting book waiting to be written for the educated lay audience about how computer systems using artificial intelligence and neural network architectures are being used to study and help understand the processes of the human brain. Such a book will help readers like myself to better grasp the workings of the mind and certain aspects of phenomenal consciousness, and will nicely complement the myriad of speculative works regarding consciousness that have been written over the past two decades by a bevy of psychologists, philosophers and neuroscientists.

Unfortunately, "How to Build a Mind" is not that book. Igor Aleksander tries to cover too many topics in too few pages, and in the end can't bring together his meanderings; he doesn't leave the reader feeling, "hey, I understand it better now". Dr. Aleksander tries to interweave an historical review of philosophical thought regarding the human mind with his own life story, and sprinkle in some details about the connectionist computer tools that he has devised to mimic certain brain functions. But he doesn't tell you enough about the philosophers and their thoughts to trigger any "ah ha" sensations. He offers some tantalizing details about what he tries to do with his computers and how they attempt to do it; but just as you start getting interested in, say, iconic learning processes, he jumps to another line on his vita, discussing another assignment at another university, somewhere in another English town.

The overall effect is, well, jumpy; he doesn't stay long enough with any one topic to leave enough "flavor" to blend with the next discussion. In the end, it's an uncooked stew; the carrots, potatoes and meat chunks are floating apart in luke-warm water. Dr. Aleksander does make one point that almost serves as a leitmotif: the power of feedback and looping processes in understanding and simulating the workings of the mind. It's a point that has been emphasized by many consciousness researchers and thinkers, e.g. Gerald Edelman and Richard Hofstadler. Instead of imagining himself having unfocused conversations with long-dead philosophers, Dr. Aleksander should have spent his writing energies considering and comparing his own work with theirs. Instead of taking Aristotle on, he might have addressed the criticism of one of his lectures by a living philosopher (Dr. Margaret Boden), in lieu of congratulating himself for being taken seriously by someone in today's "consciousness club".

Another annoying thing is Dr. Aleksander's perceived need to present his own opinions regarding what human consciousness is. His views basically amount to simplistic functionalism; but unsophisticated or not, they are really quite unnecessary. As Susan Greenfield points out in her "Private Life of the Brain", Aleksander's work, however useful, is nowhere near mimicing the extremely complex dynamics of the conscious human brain. And yet he keeps hinting that his machines are already transcending the spooky threshold between working brain tissue and subjective feelings and self-awareness, and are on the verge of answering the questions of the ages. Face it, Dr. Aleksander: you chose to be an engineer, and engineers do their best work toiling in the shadows. You are perhaps not destined to be another consciousness "rock star" like David Chalmers, Daniel Dennett, Sue Blackmore and V. S. Ramachandran. But if you keep your nose to the grindstone, it may contribute to eventual conceptual advancements that will make their present debates seem like 18th Century discussions of phrenology and cosmic ether.

I will admit, finally, that Dr. Aleksander's dream sequence regarding a shared stage discussion with the likes of Pinker, Dennett and Crick does give a very compact and incisive summary of the basic issues in the modern consciousness debate. Despite their being quite removed from the immediate import of his own (underexplained) work.

The book is a parallel presentation of the evolution and struggles of AI on one hand, and the author's personal evolution and struggles on the other. It tells the story of an experimentally-minded academic who has to balance his thirst for knowledge with personal, political and bureaucratical considerations.

I think anyone involved in so controversial a field as AI is prone to "err" into philosophy, and Aleksander's imaginary dialogues with philosophers from Aristotle to Dennett are entertaining and to the point. I'm puzzled why he seems to favor Searle over Dennett, when Searle's vague points about "aboutness" are a pale reflection of Dennett's extensive explorations of intentionality. (For those who label Dennett's approach "materialist", the paper "Real patterns" could be an eye opener.) For the non-technical level of the book, the intuitive explanations of neural networks in terms of dimples or attractors are as good as they can be. Given the author's "hardware" background (Sophia, Minerva etc.), his anti-software bias is understandable, but by the time we get to MAGNUS a strange position emerges (pun intended): On one hand he honestly accepts that MAGNUS is a software simulation, and clearly recognizes the advantage of doing it this way. On the other he completely muddles the waters when answering the question if a machine can be conscious: my impression is that he's saying that the software-MAGNUS is just a simulation we use to figure things out (and not capable of consciousness), but once we got it down we'll build a neuron-based hardware-MAGNUS which will be conscious. Huh?

The references are a good selection for those who want to study further. Just one correction: Rosenblat's book is titled "Principles of neurodynamics; perceptrons and the theory of brain mechanisms", not "Introduction ..."

Igor's passion for artificial consciousness is well known. This book talks more of philosophy than the engineering behind his passion. It has some intersting imaginary dialogues with some of the pioneers in the theory of consciousness. It stirs more questions than it answers. Nevertheless its a good read.

First, I must admit that some of what Aleksander says goes over my head, at least right now. (I'm not that seriously into this subject, at least so far.) But you can tell, just from the way he writes and what he says, that he comes from a position of genuine competence WRT his subject. Also well worth noting: Either Aleksander has exceptionally good language skills, or a great copy editor, as I saw very few errors in grammar, spelling and so on.