Customer Reviews

The Engine of Reason, The Seat of the Soul: A Philosophical Journey into the Brain

5 star:		(6)
4 star:		(5)
3 star:		(2)
2 star:		(2)
1 star:		(1)

 

 

Churchland is a great philosopher who has made many significant contributions to the study of the mind. Unfortunately, most of those contributions lie in his papers, other books, and works co-authored with his wife, Patricia Churchland. "The Engine of Reason..." is aimed for the 'popular science' crowd, and it is a wonderful introduction to vector coding and some introductory neuroscience. But it is surprisingly weak in philosophical arguments. It really reads like a light, scientific textbook, and the bulk of it consists of oversimplified explanations which rely too heavily on scientific findings that aren't thoroughly established yet. He is extremely unfair towards philosophers who aren't eliminative materialists (like Searle, Nagel, etc.), and he spends literally no time refuting their arguments. Instead he bullies the reader into believing that the above writers must hold some antiquated Cartesian view which relies too heavily on intuition. He knows he has science on his side and is rather insulting towards philosophers, making them look like idiotic armchair scientists. While unfortunately philosophers are notorious for that fault, they also ask some pretty good questions and make you think. Churchland does neither in this book. This book is a real good starter for vector coding and neuroscience. But for 'popular science' that's scientific but extremely philosophical, I haven't found anything yet that beats Daniel Dennett's Consciousness Explained. For a good refutation of Searle, Nagel and the rest, read their own works and don't just listen to the brief overview Churchland gives.

I can't evaluate the neurobiology in the book since I'm no scientist, but Churchland's entirely accessible discussions of vector coding, feed-forward and recurrent networks, and the general landscape of contemporary neuroscience were exhilarating to read. They made me want to rush out and buy textbooks on the brain--a pretty impressive achievement, as far as I'm concerned.

Churchland's philosophical perspective, as anyone familiar with his work will expect, is thoroughly naturalistic. He has very little patience with anti-reductive arguments, and the three he discusses (Nagel's, Jackson's, and Searle's) receive straw-man treatments, though like everything else in the book, each treatment is good-natured and fairly humble. Readers already lacking tolerance for Searle will enjoy Churchland's caricature of The Rediscovery of Mind as a Betty Crocker cookbook.

Though his explicit discussion of anti-reductionism is sparse, the rest of Churchland's book serves as a demonstration of how much exciting work can be done if we simply ignore armchair naysaying. So I was more bothered by his lack of engagement with philosophers already on the elimintivist bandwagon. His discussion of Dennett, in particular, was cursory and frustrating. It seems to me that he conflates Dennett's distinct accounts of consciousness and content, needlessly (and in the relevant sense inaccurately) portraying Dennett as being a friend of robust human uniqueness.

But quibbles aside, the book is a fantastic read. Its optimistic view of the possibilities of computational neuroscience is infectious. Anyone without ideological blinders on will come away excited about the future of brain research.

I've not come across a more sensible and lucidly written philosophy book. The author loves and deeply believes in science. He shows to my satisfaction that the hard sciences can answer many humanities questions or make them clearly pointless. The chapters on vector processing are still not quite as scientific as the author would like them to be, but the book overall has significantly improved my understanding and appreciation of human and mammalian minds. Since Amazon doesn't do it, here is the table of contents: (1) The little computer that could: the biological brain, (2) Sensory representation: the incredible power of vector coding, (3) Vector processing: how it works and why it is essential, (4) Artificial neural networks: imitating parts of the brain, (5) Recurrent networks: the conquest of time, (6) The neural representation of the social world, (7) The brain in trouble: cognitive dysfunction and mental illness, (8) The puzzle of consciousness, (9) Could an electronic machine be conscious?, (10) Consequences for language, science, politics and art, (11) Neurotechnology and human life. Looking at the Index at the back, the entry that occurs on the most pages in the book is "prototype" which in this book means pretty much the same as what some other authors call a paradigm.

Developments in neuroscience over the last decade will doubtless be compared to the Copernican revolution, totally changing the way educated citizens think of human nature. Churchland provides an invaluable guide to recent research into neural networks, surveying the social, moral, legal and philosophical implications of contemporary neuroscience. While not giving as much emphasis to either the emotional dimension of cognition as Damasio (DESCARTES ERROR) or its social origins and ethical function (see my BEYOND RELATIVISM), Churchland's focus on neural networks has compensating advantages, especially for anyone interested in learning how to think about the way the brain works when we think. Must reading for philosophers and social scientists who are aware that the human brain is neither a blank slate nor a serial computer

This book is the hallmark of the connectionist dream -the belief that all aspects of mind, brain and consciousness can be explained by calling up neural network models-. Now the basic premise behind all this I will not contest. The brain is a large parallel distribuited processing network of neurons. But there is another big step from this to the statement that everything the mind is is a vector coding of a neural network. This is far too siplistic. Churchland of course realizes this, but continues to talk of connectionist models like neurosciences messiah.
This is perhaps only one aspect of Churchlands book, however. Overall, the book attempts to reconcile philosophy of mind with neuroscience, and it succeeds to an extent. In many parts the discussion falls into vector coding talk, but in many others it stellarily accounts for deep problems. It is a good introducion to neuroscience, neural networks and philosophy. Churchland does not present his own strong theories, but he does well in staying away from controversy. The best part of the book is in my opinion, the attempt to build a framework of the impacts neuroscience has in social and philosophical domains. This is not done often enough, and if it is, rarely with such lucidity and clarity.
Now I would have ceritanly liked much more speculation when it commes to consciousness, given the Churchland's contribuition to the literature. But he refrains from this and merely describes some other models, like Llinas thalamic oscillations, and is content in stating that it is at leas possible to see what an explanation for consciousness would look like from a neuroscience context.
The book is a grat read, and students of philosophy, neuroscience and cognitive science should enjoy it.

The book comes in two parts. The part one, which takes up more than a half of the whole book, explains what recurrent neural networks are and how those can be used to explain our own cognitive functions. This is generally a good introduction, I think. His style is casual, and we see certain smugness you normally expect at a college lecture, e.g., introducing certain authorities as his friends and presenting the picture his own daughter and the medial and lateral brain stereographs of his wife (Patricia Churchland). Like other popular science books, however, his description of neural nets is far from precise but let's not expect too much from a book of this kind. Unlike what some of our reviewers below suggested, he minimizes the use of scientific jargons and when he use such jargons he explains what those are. The first part was overall very much enjoyable to read.

You cannot expect it to be a fully philosophical book, though. His new epistemological framework arises from this newest perspective the theory of neural networks has created. To know what neural nets are is immensely important. Let's remind ourselves of a classic work in cognitive science and neurobiology. It's David Marr's _Vision_. There Marr expresses the view that physical (hardware) implementation is quite irrelevant. Now we know this is not true. To understand why this is so one may have to consult the part one.

The problem area is the part two. The chapter 11 was full of hopes and lots of blah-blah-blah's that bore you to hell. What's interesting, and makes you slightly angry, is his explanation of consciousness. Perhaps that is because Churchland's argument seem amazingly simple. But, to think about it, it has to be simple. Otherwise it cannot be a reduction. If you want to argue against reductionism, you need to bring up some form of dualism. In fact, this is what Searle does. Searle's arguments are not directed agains neural networks. His favorate scapegoat is symbolic computation. But this is something researchers have done away with a long ago. I personally think Searle never really understood what neural nets are.

What's not really satisfactory are these: Some will find he never really defeated Nagel and Jackson. I should agree with those who think so. If ever he did, his argument lacked logical clearity or I am very dumb. He is not successful in constructing a model of consciousness, either. The problem is, he thinks he is. Like Newton did, and Euclid earlier, he tries to create a set of descriptive axioms to come to grip with consciousness. But unlike Euclid, Netwon, and Einstein (remember his two postulates), some of his axioms require a first-person perspective. (ref. pp. 213-214) For example, to verify that consciousness disappears in deep sleep, somebody obviously has to go to bed. However imprecise, MEG maybe used to detect conscious activities in a live brain. But there exists no 3rd-person method to verify consciousness is a single unified experience. Churchland has been successful in explaining a lot but I think we still have a long way to go. And his descriptive theory is not adequate.

Plus, there is a misprint in page 230 of the softcover edition. The "o%cial" should be read "official".

this is truly a fantastic book. churchland presents his ideas clearly while introducing us to fascinating new results from research with artificial neural networks, which make a strong case for his brand of materialism.

while looking at a review here and reading the back-cover i noticed a mistake people are making regarding interpretation of churchland's view. churchland is *not* a reductionist, he is an eliminationist. for churchland neural states are not the same as mental states; "mental states" as-we-know-them are part of a theoretical framework known as "folk psychology" that's doomed to disappear just as beliefs in witches, caloric fluid, and demonic sources of illness disappeared from our knowledge pool.

Paul and Patricia Churchland are the leading advocates of the ¡§it really isn¡¦t anything¡¨ branch of materialist views of consciousness. They assert that one need only investigate how the sensing, processing, and reactions are generated by our brains, and this provides a full description of consciousness. They do not use the label, but they are behaviorists.

While I consider this approach very wrongheaded, Paul Churchland¡¦s detailed presentation of this idea is an insightful work of a keen intellect, and I learned a great deal from him about brain function.

What he discusses are the characteristics of neural networks. Neural networks are distributed analog processing nets, where multiple layers of inputs, secondary, tertiary, and quadrenary nodes are linked, and the interaction strengths of these linkages are adjusted in order to perform a task. Such networks have been shown to be quicker and more accurate than algorithmic programs in doing things like facial recognition, linguistic grammar checking, and most of the other discrimination functions we do. They are also tolerant to incomplete input data, and fault tolerant to loss of some nodes and connections. Since this is how the brain is actually wired, he considers the strengths of these networks to be of tremendous relevance to brain studies ¡V much more relevant than the behavior of linear computers with CPUs.

Basic neural nets have no time-functional behavior, but he points out that if the processed states form middle and upper tiers are fed back into the net as part of the input data, then such nets become capable of periodic state behavior, such as the pattern of activation of muscles in a walking or running leg. Such cyclic behavior can be continued without new inputs, and will be changed by new inputs ¡V much like animal/human behavior. Since neural activation regions when running DO follow a cyclic pattern, and animal/human brains HAVE a huge number of such back-looping neuron connections, he also considers this significant relative to our brain function.

The alternative to materialism, spiritual dualism, he rejects out of hand with cursory dismissiveness, despite admitting that the vast majority of educated people in the world hold by it. The real target of his book is what he calls ¡§folk psychology¡¨. This is the belief that we are:
{ Self-conscious
{ Hold beliefs
{ Have emotions
{ Have the power of reason
He considers all of these assumptions to be fundamentally false, and wants to replace them with descriptions of neural net characteristics as a new ¡§folk psychology¡¨.

He proposes a list of what he considers consciousness to consist of, and then claims that recursive neural nets, combined with some other observed aspects of bran wiring, fully explain this list.:
1. Short term memory
2. Independent of Sensory Input
3. Steerable Attention
4. Alternative interpretations of ambiguous data
5. Disappears in deep sleep
6. Reappears in dreaming
7. Senses integrated in single unified experience

He claims:
1. Short-term memory is typical of feed-forward recursive neural nets
2. Recursive nets run without sensory inputs
3. Networks discriminate, and the feed-back nature of this discrimination can lead to focused attention
4. The ¡§fill-in¡¨ capability of these nets deals with ambiguous data
5. Specific types of neural activity disappear in sleep
6. And reappear in degraded form in dreaming
7. One part of the brain, the interlaminar nucleus, is linked to all the sensory regions both ways, and its feedback data is common to all so must include all sensory modes in integrated form.

I consider his explanation complete for 3 of 7 features (1, 2, 4), only partial for 3 (steerability is a feature of these nets, the source of the actual steering is not so obvious, and how steerability creates ¡§attention¡¨ is unexplained), only correlating for 5 and 6, and wrong on 7 since the links he describes have low bandwidth and seem to be a timing correlator for the brain. So his ¡§explanation¡¨ is not complete even for his own limited list of consciousness features.

Other writers have described other features they consider even more crucial in understanding consciousness. A partial list is:
8. Stream of consciousness
9. Perception being absolutely immediate, vs the weak immediacy of hypotheses and memory
10. Willful initiation of activities
11. Experience of ideas
12. Hypothesis ability (idea, create mental model of world, test in mental model)

Churchland¡¦s left these other essential features of consciousness off his list of what needs to be explained BECAUSE HIS MODEL CANNOT EXPLAIN THEM. Each is much better explained by the alternate materialist hypotheses of Daniel Dennett, Nicholas Humphries, or Karl Popper. Since he dislikes each of their views, he chooses to try to mislead his readers about the nature of the problem in order to make his own views appear more palatable.

In addition to his unsuccessful effort to explain consciousness away, his model is further subject to an explicit falsification. It is this: he asserts that consciousness IS just the feedback behavior of recursive neural net loops in our brains. But since we have millions of such recursive loops, and 99+% of their activity is totally UNCONSCIOUS to us, it should be immediately obvious to any brain researcher that this claim "consciousness IS JUST recursive neural net functionsing)is absolutely false. That Churchland makes this assertion a central feature of his views shows an almost bizarre lack of self-reflection. That this is a common flaw with all Identity Theory based materialist models is no excuse for Churchland.

Excellent book, Paul Churchland's explanation is extremely clear and he makes the subject come alive. He covers machine consciousness, brain diseases, the puzzle of human consciousness and ethics. I loved the chapter on consciousness, the author gives his take on John Searle's and Thomas Nagel's view on consciousness and also explains why the only way to go forward is by understanding more about the brain. This book should be read by anyone interested in neuroscience.

I think this book is filled with the content for a specialist cognitive oriented crowd, but written for a generalist audience. I find this a blessing as it is easy to read yet does not walk away from technical descriptions. Each chapter (and section) is highly readable keeping in mind that this is an academic text.

Basically Churchland shows how cognition can be broken down into neural correlates. I learned more than anticipated (i.e. intricacies of the human auditory system, and stereoscopic vision: and comes with lenses to try out seeing in stereo). He looks at the perceived reality/or the reality as observed from the sensory system and then explains it from possible root anatomy. Fun to think about. I think that the book is more of a reference for me as I spot read this one from time to time and have never made it from cover to cover - BUT, imagine some could.

For those interested in the philosophy of cognitive science or in cognitive science this is a good find.