Artificial Intelligence Books

This is a useless book. Their definition of software agents is vague and seemingly includes 90% of all software out there today. Technical details of any kind are not given. The book fails to discuss agent technology within its proper context of artificial intelligence, distributed computing, and communications. Terms and concepts are brought up as nothing more than buzzwords. This book is all hype and salemanship with no substance whatsoever. It reminds me of the many terrible books written 10 years ago on artificial intelligence applications that failed to convey to the reader a deeper understanding of artificial intelligence itself.

For those interested in really learning about agent technology, I recommend titles such as Multiagent Systems from MIT Press.

Yet another wordy dud on software agents. One example is worth a thousand words.

This is a valuable book especially when compared to the dry, heavy feel of a technical text of other agent books. I think that the authors have used a technique that more professionals should use, as opposed to a straight lecture style.

There are business opportunities on every page of this book.

The authors have developed an important book with immaculate detail, strong writing, an excellent grip on the material, and a strong sense of direction for the future. The book is particularly strong in its earlier section, dealing with the development of agent technology. Despite the avalanche of books written about agents in recent years, there have been comparatively few books that deliver a comprehensive account of what it all really means.

I first read this book after getting a copy from my local library. I liked so much I bought a copy. The first thing to realize is this book is a management book about Software Agents. It does a great job of explaining the various aspects of agents. It could also be used for as a college textbook. There aren't sets of problems, but there is a whole lot of basic knowledge. I don't have a lot of agent experience, but I have more now after reading this book, but I was never lost reading this book. The book is a great starting point to learn the field. When you finish the book you will have a basic understanding of the different branches of the field, and the whole book is truly top notch.

How does the development of "artificial" intelligence fit into biological evolution? George Dyson suggests that the fit is seamless. This profound investigation of the history of thinking machines and evolutionary theory is brilliant and engaging. It offers a far more palatable look at the human-machine future than the misanthropic vision of Ray Kurzweil in THE AGE OF SPIRITUAL MACHINES (Viking, 1999 ). (Palatability is no assurance of accuracy, of course, but it sure feels better going down.) Where Kurzweil sees machine intelligence as better than human and confidently predicts that we will upload ourselves and abandon our bodies, -- drawn initially by the superiority of cyber sex -- Dyson envisions a spreading macro intelligence that will involve humans in wholly new ways -- and suggests it is already emerging. He predicts that much like other life forms which share their eco-niche, we are apt to become symbionts with the machines, each doing what we do best and benefitting the other. But perhaps I have gotten ahead of myself here. Are there living, thinking, artificial minds on this planet already? Dyson asserts that it depends entirely on one's definition of "living," "thinking," "artificial," and "mind." There are self-replicating cybernetic entities evolving within computer networks. Self awareness is not yet evident, but it is not entirely clear that it doesn't exist. How will we know when and if it emerges? We don't have a clear definition of consciousness as it applies to our own condition, which makes consideration of the whole issue iffy at best. Tracing the history of evolutionary theory, following the startling course of mathematics in the past century, noting the lightning fast advances in silicon technology, and all underlain by a trenchant sense of human development, Dyson weaves a fascinating tale. He pulls in the Darwin family, Darwin's critics (then and now), oddball tinkerers, mainstream theorists, nuclear physicists, Turing, Godel, and von Neuman, and the science fiction of Olaf Stapledon in the telling. A highly rewarding if sometimes difficult read. Dyson ends with a quote from Thoreau, suggestive of his own open-ended view of our future: "In wildness is the preservation of the world." Exactly so.

EDVAC architecture by Von Newmann changed the world. Von Newmann chose to adopt the McCulloh-Pitts symbolism for diagramming logical structures of stored program codes. EDVAC had the ability to modify its own instructions similar too the theoretical Turing machine. EDVAC stored both data and instructions in mercury delay-line memory as binary and as in the Turing Universal Machine, long strings of bits represented numbers to be operated on and sequenced and potential dynamic structures of operations to be performed, such as bit shifting, multiplexing, Boolean logic, memory storage, and accumulation.

Von Newmann's next machine was called the IAS. The initial development of the IAS design was distributed to multiple locations. A central processor operating in parallel on multiple bits of a word of data at a time characterized IAS. ISA had a hierarchical memory range with random access to memory on limited media, and a distinction between software functionality and hardware functionality. "Science, as well as technology, will in the near future and in the far future turn from problems of intensity, substance, and energy to problems of structure, organization, information and control." Von Newmann was persuaded that the high-speed computer would change the nature of mathematical research. The IAS machine contained the world's first fully functional random-access memory, RAM. Disk storage was provide through 40 cylinders arranged in a bank of 20 with 1024 bits per cylinder; additionally, 40 Williams tubes and 2,600 vacuum tubes performed digital processing with a 75% up time. IAS included an arithmetic unit, accumulator, two shift registers, an adder, and a digit resolver. Floating point was considered but not implemented. IAS included 20 basic instructions and 44 order codes.

Human calculators provided the pattern of processing modeled in the computer. Human calculators demonstrated coordinated computing, sequencing, and analytical capability. Human calculators worked in parallel managed and coordinated processes deciphered WWI Germany encryption messages. The brainpower and segmented problem solving 10 X 15 power number combinations.

The human calculator model could be simulated in the Von Newman and Turing machine and the connection machine architectures and software. Neural Nets could be model in the Turing machine.

However, evolution algorithms will not be able too produce a thinking machine. Thinking is limited to the humans and divine beings. Behavior can be represented in Finite automata graphs, AFSMs, and mechanized behavior may appear logical but this does not suggest the machine can cross the sphere into human intelligence. The title of the book directly is a criticism against the evolutionary humanist. Turing grammer suggests discrete processes can be interactive described by a language. Computer automata can not evolve beyond discrete functions and the machine will be confined to the range of mathematical theorem proofs. Mathematical reasons does not encapsulate all human reasoning and such an acceptance of this conclusion would be uncreative, limiting, and lacking in vision of the potential for humans to feel love, joy, and acquire greater intelligence.

Von Newmann saw digital computers as mathematical tools, a general class of automata and did not imply they could think. Von Newman became more interested in the machine reproduction. "Every automa that can produce other automa will only be able to produce less complicated ones." Celluar Automa has yet to produce a computer brain that will function. CA algorithms surprisingly can model many patterns found in nature and physics. However, no CA has produced a grammer or graph that can be reproduced by the machine yielding an intelligence reasoning machine. Von Newmann hoped for CA salvation, "there is, however, a minimal level where this degenerative characteristic ceases to be universal. At this point automata which can reproduce themselves, or even construct higher entities are possible." Von Newmann's inspiration was not CA but VLSI. VLSI were being replicated from computer generated patterns by computer operated tool. FAB in the 20th century continued Von Newmann's aspiration and robotic automated factories suggested to a minimum degree the theory had value. Intelligence move counter to entropy and if one observes a machine producing other machines of higher construct characteristics than one would declare intelligence has been proven. In "Flesh and Machine" the Brooks suggests GA do have the ability to create simple behaviors such as locomotion, tactics, and architectural models but fail too create higher-level concepts. Abstracting and creative thought are outside the realm of the machine. Brooks suggests AI breakthrough is limted by a lack of quality software, missing laws of intelligence, slow machines, and entropy caused by a lack of young Einsteins willing to dedicate their brains to solving the AI problem.

Von Newman in his "Theory of Self replicating Automa" believed automa would grow more complicated from one generation to the next; no device would become the brain; high speed switching was millions of times faster than biological neurons but pales in comparison with the combinatorial ability of a billion neurons; and something as complicated as the brain could not be designed but had to be evolved. The idea that perfection could be reached by random arrangement of neurons seems doomed to fail. Von Newmann suggested growing a matrix of artifical neurons. These neurons should have the tendency towards self-organization among large number of interconnected secondary machines. Incomprehensible complex processes among the secondary machines could be observed by humans have the appearance of comprehensible behavior. Brooks simple behavior modeling through AFSM seems too synchronized within the realm of computer theory. Imitation verses enhancement feels like imitation is more discrete, definable, and programmable. Enhancement seems to be the result of complicated imitation and the AI is the failure to adequately define AFSMs too model observable behavior. AI evolution must be confined to the realm of the Turing machine grammar. To expect a machine to suddenly start thinking and its neurons to behavior like biological counterparts is a myth, a fable to consume brilliant minds into the dream that machines can think.

Though well written and informative, in the end DAM was a less than satisying read. Dyson marshals considerable data (and extensive and informative quotes) from the fields of history (of science and technology), the sciences (principally evolution and CS), and philosophy (as it has, historically, reflected on notions of mind and evolution). As an avid reader of history, with a deep interest in all of these subjects, I found the opening chapters of DAM quite interesting. That said, the history in DAM is not particularly deep. But Dyson writes well, and I appreciate his having shed light on several lesser known (and underappreciated) historical figures along the way.

Where DAM ultimately falters, in my view, is in its shallow futurism. I say "shallow" not because I don't think Dyson is highly imaginative. He is. And his predictions (to the extent he articulates them as such) may well be realized one day. However, though Dyson is skilfull in establishing the historical groundwork for the development of computer and communications technology as they exist today, he is far less skilfull in tracing even a speculative chain of developments from the present state of the art to the global/artificial intelligence he envisions as a possible (perhaps inevitable) future development. In fairness, every futurist has hit and will continue to hit this wall until the future comes knocking. But Dyson purports to do so.

In the final analysis, though Dyson does an admirable (and entertaining) job of accounting for the rise of computers, and the increasing complexity of computer networks, his discussion of artificial intelligence has more the ring of a leap of faith. It's a fascinating idea (though hardly original to Dyson), and certainly a possibility, but one whose potential trajectory (from idea to realization) is barely even attempted in DAM. DAM would have profited from a little more hard science, and a little less soft speculation.

Several have criticized Dyson's philosophical and historical treatise "Darwin Among the Machines" for not articulating exactly how a global intelligence might emerge from today's synthetic biological and computational networks. But as Dyson says in the preface, the past is where we find answers, and the future merely a fog of questions "to which the answers are up to us." In the next 200 pages, Dyson explores the history of an idea: that man will someday create a form of artificial life, with intelligence that may match or exceed our own.

It may astound some readers to know that these ideas date much farther back than Alan Turing's "Turing Test," or Vannevar Bush's influential essay "As We May Think." Consider the following quote from Thomas Hobbes (1651): "Nature is by the Art of man, as in many other things, so in this also imitated, that it can make an Artificial Animal." Or consider this excerpt from Samuel Butler's 1859 essay, which serves as Dyson's main theoretical foundation: "As the vegetable kingdom was slowly developed from the mineral, and as in like manner the animal supervened upon the vegetable, so now in these last few ages an entirely new kindgom has sprung up ... It appears to us that we are ourselves creating our own successors."

Careful to acknowledge his predecessors, Dyson profiles the lives of some of the most prescient Enlightenment- and modern-era thinkers in captivating detail. In so doing, he traces the evolution of the "Artificial Animal" from its earliest incorporeal appearances - as merely an idea - to its current computational incarnation in neural networks. But Dyson doesn't stop there.

In fact, he goes on to argue that the global telecommunications network (primarily the internet) may provide the appropriate architecture for a kind of global, distributed intelligence to evolve. Here Dyson borrows from Leibniz, who noted that the "soul" may be "born when the machine is organized to receive it, as organ-pipes are adjusted to receive the general wind."

To further support this claim, Dyson draws parallels between the development of increasingly efficient machines and the processes of biological evolution. In fact, this is one of the most interesting parts of the book, in part because the language in which Dyson details the principles of evolution might be considered dangerous today, in the midst of the raging Intelligent Design debate. For example, Dyson suggests that evolution itself may embody a kind of intelligence, though we frequently perceive it as merely a shallow process, highly dependent on chance and randomness.

As Dyson points out, this perception gets to a fundamental semantic confusion surrounding "intelligence," a phenomenon well known to AI researchers in which problems once thought to require intelligence are then seen as trivial after an algorithm is designed to solve them. As Dyson points out, intelligence may simply be a word we use to describe behavior that corresponds to our view of how humans behave. Not believing in "'the existence of an intelligence behind the achievements in biological evolution may prove to be one of the most spectacular examples of the kind of misunderstandings which may arise before two alien forms of intelligence become aware of one another.' Likewise, to conclude from the failure of individual machines to act intelligently that machines are not intelligent may represent a spectacular misunderstanding of the nature of intelligence among machines."

Ultimately, whether you agree with Dyson's perspective is besides the point. This is not a scientific book; many of the ideas are purely philosophical, and the logic used to support Dyson's assertions frequently rests on historical anecdote and analogy. These should not be considered weaknesses, however. The real, lasting value of "Darwin Among the Machines" is Dysons's imaginative and graceful writing, his impeccable historical research, and the conceptual ease with which he integrates ideas from ballistics, biology, hydrodynamics, set theory, Cybernetics, and uncountably more esoteric subjects.

Though I won't dispute that many of these exciting ideas are far-fetched, Dyson has found powerful allies for his assertions, from Hobbes and Leibniz to Goedel and Von Neumann. So if you find yourself believing - or simply wanting to believe - in these groundbreaking ideas, then you're in fine company.

I bought this book in the hope of reading some intelligent speculations by the author about evolution, machines, and AI, which is what the title suggested I would find. However, it turned out to be a history of the evolution of computers with old speculations from the computer pioneers concerning the evolution of computers injected along the way. To be fair, the author does have an overarching thesis that he tries to weave into the historical narrative whenever some past speculation seems to lend it some support. It is that the World Wide Web - that well known network of millions of computers - may some day, at a certain critical size and running who knows what software (certainly not the author) will become intelligent in some way (also not specified by the author). Come to think of it, I think the author has used the historical angle of the book - the similar speculations of the computer pioneers of the past - as a device to lend credence to his thesis - a kind of proof by consensus. I remain unconvinced, however. His arguments (where there were any; it was hard to tell his arguments from narrative) were very weak and unconvincing. To his credit, the author did a tremendous job of scholarship for the historical side of the book. However, he left the speculative side undeveloped (at the most weakly developed) and, therefore, the book was unappetizing to me.

After reading this, I now understand the behavior of my robot vacuum cleaner Roomba a bit better. I was pleased to see that the final chapter covered neuroprosthetics and the possibility of physical immortality in the future. He closes with "The distinction between us and robots is going to disappear."

Implicit belief

There are many great books about AI.
Researches give us insight in they work and philosophy.

Any way most of them didn't understand the simple fact:

There is no intelligence without subjectivity!

And more, the Consciousness is an invented, in technical sense, fantastic phenomenon. It is speculatively allocated with various properties, but to define it in scientifically consistent way is impossible. However, the problem is not only about words. The appeal to Consciousness as a real-word phenomenon, inevitably leads to necessity to place inside the brain the central managing body that contradicts the validity. Last significant attempt, known to me, was an attempt to design the elements of Consciousness systems based on nano-tubular structure theory. Irrespective of the received results connection between nano- tubular and Consciousness has not been proven and could not to be. Nevertheless, attempts to create Consciousness systems still continue.

Phenomenon of our behavior, in general, is determined by a set of acceptable to us models of it. These models are rather dynamic and determine all sides of life for reasonable systems. They control behavior of our body in all details, and all displays of our intelligence. By way of determining models of behavior, the managing body aspires to reception of the greatest possible expected positive reaction on the part of itself, and other reasonable systems, if social.
All given above is possible to summarize in the following statement:

"The system is reasonable, if it is capable to determine it's own behavior, be guided by it's own, subjective representation about the World known to it."

That definition is actually partial blueprint for that kind of system.

It is easy to beat up my opinion about Consciousness. One should just find the hard facts that could support common implicit belief in existence of the Consciousness as a real-word phenomenon.

Sincerely Michael Zeldich

When I picked this book up, I was thinking that it might be one of those books that laud the virtues of science and suggest a coming utopian dream where robots and technology solve all of our problems. "Flesh and Machines" is not that book.

Brooks kicks off with a brief overview of the history of artificial creatures, which was particularly fascinating. It was especially interesting to learn of a 17th century duck that could even go to the toilet. The history that has gone before modern robots is one of which I have been singularly ignorant, and thus this was a deeply informative section.

Brooks does not give wildly optimistic predictions, recognising the inherent dangers of doing so. Rather, he makes conservative estimates based on what research and technology is available. It might not be as exciting, but it is certainly more realistic. Brooks also discusses the possibility of building machines that are truly conscious, which brings me to the one criticism I have.

In discussing the idea of whether humans are special or just "machines", Brooks slams three scholars for appealing to something that cannot be seen, something "other" than the physical make up of humans, and also for having no data to support these ideas. In particular, Searles comes in for some especially harsh criticisms. In the very next chapter, Brooks is doing exactly the same thing with his ideas on what he calls "the juice". Brooks is openly aware of the hypocritical nature of his comments, and this was something that I found particularly odious.

The book is quite heavy with accounts of Brooks' own work, and one can get a feeling that Brooks is also defending and promoting his research and company, (iRobot Corporation).

Overall, the book serves as a good introduction to the field of artificial intelligence, and the impact of robots on our lives. I enjoyed reading it, and thought it was a good book. It avoids overly exaggerated predictions, giving some realistic thoughts on how robots will change life in the immediate future. Brooks never really predicts beyond 2020 all that much, and when he does, it is in very vague terms. As he says, he is aware of the dangers of predicting too far and too much about the future that no one can really know about. I enjoyed the book, but I think it is more for the casual reader who is curious about robots, rather than anyone with deeper knowledge in the field.

Ghenhis was the name of a robot that could walk over anything in its path as it followed a person. Ghenhis had six legs, bumper antennas, and infrared sensors for following the heat of the person signature it is following. The software for Genghis was not organized as a single program but fifty-one parallel programs and Brooks called these programs Augmented finite-state machines (AFSMs). AFSMs can send numbers to components on fixed wires. The first forty eight AFSMs allow Genghis to scramble around rough terrain. The walk machine uses six outputs that sequence the six legs to take steps. Ghenhis allow the robot too be out of balance as it walked and the beta balance machine corrected The Infrared Sensors machine receives input from six pyroelectronic sensors and each one has an on or off state that feed into the prowl or steer machine. These sensors are motion detectors and are tuned to the radiation band emitted by mammals. The prowl machine is connected to inhibit the outputs of the walk finite-state machine. If the robot detected some infrared activity, it walked toward it. If the sensors had been rotated to the back of the robot, it would walk away. Ghenhis had no sense of the directions backward, or forward, or away; it only had interaction embedded in it according to sensor input. Steer machine had left and right IR sensors and if the left sensor fired then the machine sent out a message to the left legs to take smaller steps, and if the right sensor fired then make the right legs take smaller steps. The insectlike Ghenhis was a turning point for robotics. The Ghenhis followed an emergent trajectory that was a product of both of its actions and its situation in the terrain of the world.

Brooks robots response to situation with conditional reactions and Cynthia Breazeal set out too write AFSMs in a higher-level language called the Behavior language. Colin Angle and Cynthia Breazeal built twin robots Attila and Hannibal each with 19 motors, 11 onboard computers, and hundreds of sensors. Eventual Breazeal produced over 1,500 AFSMs with her Behavior language code and through a model of pain through inconsistent sensor readings, they were able to ignore bad sensors and reintegrate them once they started to operate again. The legs of the robot were able to cooperate when the robots encountered rough terrain, lifting the body together, holding things up while a leg search for a difficult foot holding, and backing up and going around obstacles when needed. These robots were built from layered control systems without a central cognition box and coupled sensors to actuators.

The philosophers George Lakoff and Mark John argued that higher-level representation of language and thought are based on metaphors for our bodily interactions with the world. Metaphors develop from childhood from physical and social experiences, for example affection uses warmeth because the child is exposed to the warmeth of the parents body. High level concepts are built on metaphors and rely on bodily experience in the world. Our language reflects these metaphors.

Metaphors make it worth exploring the building of a robot with a human form and seeing what metaphors can be derived from the experience. Robots are not people. However, people will know how to interact with robots in human form by making eye contact, nods, and other sublinguistic murmurs and other social clues. The robot will know when to talk and when to listen dependant on the social clues and Cog would pave the way in this research.

One way to build a robot that can interact with people is a natural way is to build it with a vision system and with eyes that saccade and verge, and that look like human eyes. Each of Cog's eyes has two cameras. One has a wide angle lens so Cog can see peripheral view and the other has a narrow-angle lens to give Cog a fovea. Each of Cog's camera eyes are mounted on gimbals that can pan and tilt and its head and neck give it more freedom of motion of exploring. When Cog looks off in a direction, its head also turns in the direction. Cog vestibular-occular reflex allows its eye motions to successfully saccade and Cog is able to smooth flow someone walking in front of it. Cog's head has a gyroscope too play the role of an inner ear.

Cynthia Breazeals robot Kismet paid attention to three sorts of things: moving things, things with saturated colors, and things with skin colors. Kismet has internal drives that get larger and larger unless they are satiated. As these drives get larger they release certain behaviors. If Kismet bored drive get large, it might start deliberately looking around, saccading from place to place looking for something. The weighting on its attention system on saturated colors will direct the eyes while saccading to bright colors in the periphery view. The overall behaviors emerge from the interactions of the simplier behaviors.

Kismet has an auditory system and analyzes four pitch types known as prosody. Human infants recognize approval, prohibition, attention-getting, and soothing through prosodic patterns. Kismet has three emotional states: its valence, it arousal, and its stance. Valence is a measure of its happiness, and its arousal is how tired versus how stimulated Kismet is, and stance is how open it is too stimuli. It displays its emotional states with a set of eyebrows, its lips, and its ears and can put prosody in its voice.

Ritchie says to Kismet, "I want to show you this watch my girlfriend gave me." Kismet dutifully looks at the watch. Kismet was picking up on the social clues and the directions of attention. When Ritchie brought the watch into Kismets center of view, a few inches below his face where Kismet was foveated and when he brought his index finger up and tapped the watch the motion actived Kismets attention system and Kismet maintained eye contact with the watch. Eventually, Kismets attention system decided Ritchie's face was more interesting and looked back at the eyes of Ritchie. There is nothing qualitatively different from the mechanism in Ghenhis.

Having read this book it's plain to see how cutthroat the world of AI research and development is. For author and certified 'AI authority' Rodney Brooks, it seems like the odds are stacked against him in his quest to bring true AI to life. He kicks things off with a mini history lesson, briefly covering everything from the 18th century automotans to early 20th century computing. His most interesting and relevant example is W. Grey Walter's 'mock turtle' toybots. He illustrates how a simple design with on-the-fly decision processing could fool people into believing that a machine could in fact possess intelligence. Running under the creedo "fast, cheap and out of control" Brooks has presented a view of AI which flies in the face of academic models. One of the more fascinating aspects of the book is how people in this field obtain funding. It seems that practicality is often overlooked in order to secure bigger budgets extending research. Brooks spends considerable time pointing out the faults of this system while patting himself on the back for trying something different. That's all fine and well - until we hit the juice.

At some point in this book the focus shifts from practical applications to the prospect of "conscious robots". After rambling on about the philisophical implications this might entail he takes a few shots at contemporaries like Ray Kurzweil and John Searle. And this is where we first learn of "the juice"; the new "stuff" that we haven't yet discovered which contains the meaning of life, the universe, everything. Considering this explanation comes on the heels of a rant against people who "think they're special" it's a little bewildering. At least someone like Kurzweil has a sliver of theory to base his wild ideas off of. Brooks on the other hand goes off the deep end with his juice theory without even the slightest argument.

It would be nice to read a book of this nature which actually addresses how an AI might actually approach something subjectively. How might an AI develop tastes, emotions, desires and impulses? With the juice? You've got to be kidding me.

In contrast with many other books on AI and neural networks this book DOES give the following:

* limited amount of topics that are explained well.
* good examples.
* well written.
* useful tips and guidelines.
* source code.

I my opinion this book does a good job of giving an excellent explanation of the selected topics. In contrast to other authors who try to show off all their wisdom by touching all possible topics in AI and then off course stay way to general in all of them this book delivers what is says.

One small point of critique: it can use a little restyling :-)

I am just starting to learn neural networks, and this book hit the spot. It is not overloaded with math, and its approach is very practical and well suited for programmers/engineers that want to start working with neural networks. I would recommend this book to anyone that has a background in computer science or engineering and wants to learn how to apply neural networks.

This book isn't for a beginner. There're many good beginner's book for NN (or general AI), but this book delves into a bit more detail, and requires careful reading and understanding to appreciate the book.

I have given negative review for this book in the past. What I want at that time was a quick reading and code. This book is not like that. You have to have some basic knowledge in AI (at least read any introductory book once) to understand it.

I highly recommend this book for any one just got bored with introductory books on AI.

I bought my first copy of this book over two years ago. I wore it out, and I just ordered a new copy to replace it. The book is just what it says: practical, step-by-step instructions on designing and training neural nets. The author gives excellent descriptions of neural nets in plain English. He explains what they can do, what the different structures are, how to construct training sets, avoid local minima, interpret weights and evaluate performance.

An earlier reviewer claims that the code is incomplete. I am puzzled by this assertion because I had little difficulty getting it to compile and have achieved some very promising results.

I highly recommend it!

As an undergraduate Math and Computing student, when I took up a high-level course in Neural Networks as an open elective, the only book my instructor recommended was Neural Networks by Simon Haykin. I got that book, and was soon put off by the mathematical rigour it had.
It was sometime later that I came across Practical Neural Network Recipes in C++ by Masters'. This, by all standards, is an exceptionally well written book.
It has the complete code for a neural network application, including Conjugate Gradient based back-propagation, Simulated Annealing and Genetic Algorithm powered optimisation, and much more. The code, although not very object-oriented, is clear and easy to follow. Undergraduates with a limited knowledge of mathematics will most certainly appreciate the way Masters' deals with the underlying concepts behind neural networks training and use. He simplifies the mathematical equations, and the code listings serve to see the math in action. The more mathematically mature can look into the excellent references provided in the text.
When much later in the course I went on to study Recurrent Networks (RNNs, which Masters' doesn't cover in his book), I found myself going back to Masters' when I had to implement algorithms for RNN training. This is one book that will teach you to convert complex mathematical equations into working code. Its a skill that is of much importance to most computational science students. This book is a must have for all neural networks students and practitioners alike.

Exploratory spaceships from Earth are crewed by digitised personalities. 110 years after it left, one such ship encounters alien ships.

These almost miraculously build a bunch of orbital towers in space. Some of the crew inhabit android bodies to go and investigate.

It turns out that these amazing alien devices have been left so that only one man can use them, probably the most dysfunctional personality they have.

One of the uses for this technology is FTL travel, so the ship can return to Earth. Their planet has been decimated by AI and not many humans are left.

Most of those that are have taken on some posthuman traits themselves.

_Echoes of Earth_ by Sean Williams and Shane Dix in the opening pages introduces us to the character Peter Alander, a member of the deep space survey ship _Frank Tipler_. Launched in the year 2050 and traveling at 80% the speed of light, the _Tipler_ carried Alander and the rest of his crew 72 light years from Earth to the star Upsilon Aquarius. They were sent there to survey the third planet in the system, a planet named Adrasteia, one that was known to be the source of oxygen and water spectra detected by interferometers in the Sol system. The _Tipler_ was one of a thousand other similar survey ships launched as part of the United Near-Earth Stellar Survey Program (UNESSPRO) in a great wave of exploration.

In a real twist Alander is not human. Nor is anyone else on the _Tipler_. Or at least, not completely human. Alander, mission leader Caryl Hatzis, and the dozens of other specialists on the _Tipler_ are engrams, electronic simulations of actual people, people that were chosen in the 2040s to have their thought processes copied and molded into a highly complex simulations, ones that could perform - hopefully - just as well as their originals could. Carefully modeled to experience such hormonally influenced traits as desire, hunger, and fatigue, they nevertheless were much better suited to the rigors of interstellar travel. Not only weighing virtually nothing (as it is extremely expensive to accelerate anything to near light speeds), they are much better suited to survive the nearly one hundred year trip to Upsilon Aquarius, able to adjust their internal clocks or even shut themselves off to avoid consumption of resources and to avoid boredom. Once at the planet they generally stayed purely electronic, interacting in a virtual ship-wide environment called ConSense, though they can elect if resources permit to grow and inhabit partially biological androids. In the case of the _Tipler_ mission Alander was the only individual to inhabit such a body, though this was done in large part to keep his engram from breaking down, as he was in danger of merging with and losing his distinctiveness within ConSense due to an apparent flaw in his pattern.

At first the survey looked routine, as while Adrasteia had life, there wasn't much of it; just simple cyanobacteria analogues surviving in the clouds thanks to the planet's bizarre climate. Things did not stay that way for long. After some unusual energy discharges near Adrasteia, strange, perhaps alien objects appear in orbit around the planet. Quickly dubbed Spinners by the crew, they within less than a day construct ten orbital towers around the planet's equator, towering structures that reach from the planet's surface up into space, connected by a vast ring. Ignoring all hails from the _Tipler_, there is only one apparent means of interaction; what looks to be an elevator of sorts in the base of one of the towers. Against the wishes and at first without the knowledge of Hatzis and the others, Alander steals a shuttle and flies to that tower and takes the lift. Once inside, he finds that the towers were indeed built by an alien intelligence, a race that is vastly superior technologically to the one the engrams came from. Greeting Alander is a highly complex artificial intelligence that they dub the Gifts, for they inform Alander that they are in fact a gift from this race, the Spinners. Refusing to answer any questions about the Spinners themselves - claiming that they do not possess such knowledge - the inform Alander that he has been specially selected to act an emissary to those on the _Tipler_, that the Gifts will speak to him and him alone, guiding him to the various treasures in the towers.

For purposes of the story the two most significant of the various gifts are faster-than-light communication, and a small faster-than-light ship that the Gifts tell Alander is called a hole ship. Using the communicator, they send a message to Earth. They receive no reply. The _Tipler_ had lost contact with Earth during the flight to Upsilon Aquarius but felt it was possibly due to any of a number of reasons, including time lag. After some discussion and a near disastrous crisis, Alander decides to pilot the hole ship back to the Sol system (being the only one that the Gifts will allow to do so).

What he finds upon his return astounds him. Though he thought he was prepared to have found humanity to have vastly advanced during the hundred years he had been gone, Alander was shocked. Venus and Earth were missing, and there was a vast structure in its place instead, the start of a Dyson sphere. Artificial intelligences, having become self-aware after the _Tipler_ left, nearly wiped out humanity, with only 3 million still alive, many of those not entirely biological any more. Alander, ignorant of current Sol politics, tries to contact whatever successor exists to UNESSPRO. Instead he comes to the attention of the Vincula, a group mind of sorts that rules the system now, one that attempts to use Caryl Hatzis, the sole surviving engram contributor, to gain control of Alander and gain access to the Gifts.

Alander's arrival touched off a conflict within the system, the Vincula viewing his apparently faster-than-light ship as extremely valuable while the few remaining individualists were not keen to let the Vincula control the technology, the latter trying to oppose the Vincula. Though that seems bad enough at first, a far, far worse conflict occurs. Humanity learns the terrible, awful consequences that result from the use of the Gifts, not one coming from the Spinners, but from some other, darker alien menace.

According to the Gifts, "There are civilizations who take delight in the destruction of others."

And thus begins a trilogy. Things look extremely bleak at the end, but there is a ray of hope. I found it very well written and highly engaging and look forward to other books in the series.

This is the first volume in a series? Trilogy? I dunno. I can say that at least two more books follow it.

So once again, it's the future: 2165 or around about that. It appears that by 2050, Earth had become all peaceable and stuff and also monstrously prosperous, thanks to technology. So everyone became real keen on exploring space. 'Cept that it would be really expensive and not terribly feasible to send human crews blasting around for hundreds of years to reach our nearest neighbors. So engram crews were sent instead: super-complex software recreations of actual people, or bodiless clones, if you will. This meant that the ships just basically had to be flying computers with some nanofacturing capabilities to build stuff at the destination. Also the engrams could basically ride along in stand-by mode, more or less sleeping, so as to not, you know, flip out through the sheer boredom of the long voyage.

Well, at this here one distant destination, many light years away, and a hundred years after launch time, one engram does wig out over the basic disconnect over "my memories tell me I am Peter but really I know I am a computer program in a VR environment". So his crew dumps him in an android body on the planet's surface and tells him to just kind of putter about at the base camp there and stay out of their way. They get no transmissions from Earth, so obviously something happened during the trip and the home planet cannot or will not talk to them (although of course any real-time communications would be out of the question due to the years-long time lag).

A coupla years later, the engrams are just minding their business and building robo-facilities and exploring and stuff, when, within a day, a bunch of linked orbital towers get connected via space elevator to the surface. Who built these, and how and why, are mysteries. Pete the engram/android flies over to the base of one of the tower-things and gets a free ride up to the spindle attached above, way up in orbit. Then a pack of alien AIs go all, "I am for you, Peter" and tell him, yeah, some benevolent super-aliens just did a quick fly-by and built this whole complex installation with some of their Model T-level technology, 'cuz they're all hyper-advanced but they like to throw a few crumbs at the more primitive species they encounter, to help 'em bootstrap their way up. And oh, yeah, the alien AIs will only talk to and obey Peter and no one else in the crew.

So the novel goes from there. Who are these aliens? What do they want? Are they good? Are they bad? Should the engrammites use all of the kewl toys the aliens have given them? And what has become of Earth in the meantime?

This is a tale on yer epic Clarkean scale with a bit of Vernor Vinge thrown in. Huge revelations are...um...revealed. And action takes place on literally a stellar level. Lots of big ideas get thrown around. (The authors are a little too proud of their use of the revised Planckian measurement system, but it shows how seriously they take some of their scientific gimcrackery.)

It's pretty good and definitely bold. Zesty, with a big finish and a slightly nutty aftertaste. I enjoyed it, and my cat Mr. Hate gives it his highest recommendation of "I would sleep on top of that book".

A engram crew of a survey ship runs into an alien race that likes to leave gifts. This gifts take many forms, mostly in the form of information or high-tech wonders. But the aliens don't stick around, but just dump the gifts and leave.
Can they be trusted? Or are they just paranoid? Do they have REASON to be paranoid?
The gifts, from a faster-than-light ship to a library full of information about the galaxy, all seem too perfect. Was the survey ship just at the right place at the right time, or is there something more happening?
This is a hard science fiction adventure, a first contact novel and a dangerous mystery all in one. Can Peter Alander figure out what to do? Can he help humanity, made up of people who look at him as nothing more then a flawed program, or will he end up failing it?
I enjoyed the novel very much and really found the idea that most of the characters are not even living beings, if defined by our standards, to be a nice touch.

Maybe it's just me, but I found this book hard to really like. It has some good elements: the aliens are intriguing and so is the technology, both human and alien. The premise of humans exploring beyond the solar system, encountering technologically superior aliens, and dealing with the consequences of such an encounter all appealed to me. Nor can I complain about the pace of the book. Events moved along and there was enough action to keep me involved. On the other hand, the story always had something of a cold, impersonal feel for me. I found it impossible to really care about the principal characters, most of whom are computer "engrams" (i.e. programmed personalities based on real people who remained back on Earth). These engrams run the starships while "inhabiting" a virtual environment within the starship computers. A "dead" engram is just a deleted program when all is said and done, and I just couldn't get emotionally involved with that. Further, what is done to humanity and to Earth, both by aliens and by ourselves and our own technology, felt both far-fetched and improbably grim to me.

I read this book all the way through but, while it was interesting, I can't say that I liked it very much by the time I got to the end. Intellectually stimulating perhaps, but not emotionally satisfying. Some readers will like it a lot, I'm sure, but I had a very mixed reaction to it. At this point, I'm not sure if I will read the next book in this series or not. I can't give ECHOES OF EARTH a strong recommendation. Proceed at your own risk.

I enjoyed Hacker and the Ants more than any other Rucker book I've read. When working in silicon valley in the 90's I lived near White Road, in the dry east hills of San Jose, and also had my share of lusting after Vietnamese girls. Reading this book in 2008 takes me back to those days and resonates with the feel of the whole place. Rucker is right on with this one.

But the hero seems "ethically challenged"...?
This cyber da Vinci is a software developing genius
but takes a fall at his bosses wishes.
At lot of times Rudy Rucker is on the money in his
futures and he seems in 1994 to see Silly Valley today
better than Steve Jobs does? AI hasn't quite kept up, but
virus technology hasn't made him a liar either.
Rudy Rucker earns his sci fi bucks the hard way.

While most definitely not your garden-variety science fiction, The Hacker and the Ants looks positively pedestrian when compared to some of Rucker's earlier work (e.g. White Light). This is the story of Jerzy Rugby, (a nod to Stephenson's "Hiro Protagonist"?), a computer programmer who has been ruthlessly downtrodden by life. When one of Jerzy's software creations goes horribly (and absurdly) awry, we are swept away in a bizarre plot that offers up intense romance, hideous conspiracy, and a pet robot named Studly. What else could a reader ask for?

Rucker skillfully mixes the real and the surreal, to create an interesting hybrid. The overt and persistent normal-guyness of Jerzy's personality cast against the array of strange happenings adds a nightmarish tint to the story; this is an average man trapped in an increasingly absurd and hallucinatory narrative from which he cannot wake. At points, however, the story goes a little too far over the top (for instance, with names like Jerzy Rugby, Bety Byte, and Krystal Kattle) exposing the seams of the novel.

That is, of course, the ultimate shortcoming (or genius, depending on your point of view) of all of Rudy Rucker's work. Just as obviously, this reminds us that Rucker is not for everyone - for those who are more straightforward thinkers, this sort of style is an acquired taste, at best. But, if you are one of those who enjoy the absurd and at-least-slightly-surreal, I'd recommend the book (and the author) strongly.

In this book, perhaps more than his others, Rudy Rucker takes the common themes of Cyberpunk literature (the frightening/enlightening possibilities of technology) to unique and fascinating ends. For those looking for a taste of just-over-the-horizon technology in an all too familiar environment, you will find much to like here. There is the dependence on visual stimulation and passive entertainment, whose destruction brings about blood-thirsty mobs. There is the shift away from real-world social interaction and towards anonymous techno-societies that makes communication increasing awkward when not accompanied by digital interfaces. Finally there are the ethical and logistical problems involved in giving increasing control and autonomy over the machines that do not necessarily share our moral qualms.

In the end, however, the writing fails to live up to the ideas. The first hundred pages are awkward at times, excruciating at others. Once the meat of the plot begins, the overall writing seems to improve, but the dialog still seems stilted and the pacing is jumpy. Although I had hoped to attribute these factors to the stylistic choices made by many modern sci-fi writers, it became obviously that the writing simply wasn't up to par.

If you want high quality Rucker, you will be better off with his later *ware tetrology.

I'll start by saying what's wrong with this book. Rucker must have had a really bad marriage as he uses his wife in his books rather negatively. But even with this he writes a fantastic book. I can look at the technology that he describes and almost see it in use today. No magic but solid tech. And the story fits so well with the genre of both cyberpunk and today's news. I really wish this book was in print for others to read.

Paranoia, with clones.


Just no color levels.

A rather different book than the first in this series. Spin State had a heavy focus on the quantum. Spin Control has a heavy focus on information, and problems with complex systems.

These include spies, so a lot of distrustful spooks hurling accusations (or weapon fire) at each other.

There's a colony to worry about, a killer virus, and more.

Li, the focus of the first book, only appears in parts, so big fans of hers are likely to be a little disappointed, I think.

The spy story thread focuses on the situation in Israel (and, of course, Palestine), where people have been trying to fight wars with AI. Apparently AI minds suffer from the mental problems associated with war even more severely than people, so this particular subplot is the most interesting part of the book - coming to the fore towards the latter, and strongest part of the novel.

A decent book, but not as good as the first one.

The events at the end perhaps leave this open for Moriarty to take a jump ahead in time in this future history she is sketching with another book.


3.5 out of 5

Spin State (the first novel) was so good I expected the typical slump with the second novel but it was anything but. This is science fiction for 21st century grown-up's who apprecate a richer, deeper and more literate experience.

Reading the description or the back cover you'd be forgiven for thinking "I've read all this stuff before" but you'd be dead wrong. Where so many authors attempt and fall short, Moriarity turns all the dials up to 11 while simultaneously breathing fresh air into so many sub genres.

The main characters are all interesting with unique motivation and depth and there are more interesting ideas per page than most authors writing today, something which reminds me in an odd way of what Phillip K. Dick might have been like if he were writing today without the drugs. :)

There is no question that Chris Moriarty can write. She wields simile like a master swordsman, and her scenes read with the detail of an author who has really thought them through--has visualized them from every angle. Her story devices also required skill. Her first book, "Spin State", was a mystery novel set in a speculative future while this second book, "Spin Control", is a spy/intrigue novel set a few years after that. With all of these demonstrated skills, why do Moriarty's books read like someone post-holing through deep snow?

I think that, perhaps, Moriarty has been a bit too ambitious with her material. I count four main topics, Space Opera, Science Fiction, Cyber Fiction, and Action/Adventure, all crammed together and competing for space on the page. Any quarter of this book, if developed completely, would have been a well-drawn, fascinating story. Taken as a whole, though, each topic is crowded into a confusing jumble of undeveloped characters, confused motivations, and shaky chapter transitions. In my opinion, the storytelling in "Spin Control" does improve over "Spin State", and Moriarty's ideas remain fascinating albeit tedious in their development. Perhaps more improvement will come with a third book.

After enjoying Spin State so much and seeing many positive reviews by other Amazon readers, I was really looking forward to reading Spin Control. What a disappointment! I knew that it wasn't going to be a sequel per se, but I did expect the same interesting plot, characters and social/technical context that Spin State offered. Instead I found a very confusing plot (I'm still trying to figure out what happened and why, a couple of days after finishing the book) plus a lot of talky lectures about complex systems that left me bored and further confused. I do agree with other reviewers that the characterization is better in Spin Control and the inside-Syndicates perspective provides an interesting counterpoint, but these factors weren't enough to redeem the novel for me.

This is the first Amazon review that I've written, motivated by the need to give a minority report balancing all of the five-star ratings found here. I'm happy that those reviewers loved the book, but personally I wish I had quit reading Chris Moriarty after his excellent Spin State.

There are parts of this book, particularly the last 100 pages, that are hard to put down and will keep you reading through phone calls and well past bedtime.

Unfortunately, there are parts of the book that are also weak...or so fascinating that just by touching upon them rather than exploring them more...I was left unsatisfied and a bit perplexed. I agree with several of the reviewers that the parts of the book about the Syndicates were fascinating and desperately needed to be explored far more.

Also, the plot is almost ludicrously convoluted and the chart the characters draw up to get their own hands around the plot made me think first, Wow, neat idea, and then, Wait a minute...this is what's wrong, when's the last time I read a great book that had to include a chart of various characters and plotlines to keep me focused...? Never, that's when, and that's the fundamental problem.

The characterization is excellent, the setting perfectly spun (albeit depressing), and the writing top notch. Unfortunately, the plot spins a bit too much out of control, is too convoluted and thus ends up losing a lot of its punch...which then caused my enjoyment of the book to drop.

If you like Moriarity's earlier book, SPIN STATE, this is an enjoyable, if not great read. If this is your first look at Moriarity or this type of cyberpunk sci-fi, there are a lot better books out there to whet your appetite on.

I'm really interested in machine learning and artificial intelligence, but the author's outright assumption that he is going to do for thought what Schrödinger's did for DNA (before its discover) is pretty arrogant. I don't know what it is (perhaps using mind as a proper noun inconsistently), but halfway through the book I couldn't take it anymore and ditched it in favor for Hawkins' "On Intelligence". Life is too short.

In the introduction to this handsomely bound book, the author suggests that it is an appropriate time for an explanation of how the dynamics of a human brain can be accounted for by computer science. His title is motivated by Erwin Schrödinger's enormously influential "What is life?" which launched the field of evolutionary biology by inducing both Francis Crick and James Watson to successfully seek the molecular basis of biological evolution, but the analogy is strained for several reasons.

Schrödinger's book is less than 100 pages in a current edition, while Baum's is about five times as long. In the context of Schrödinger's lifelong interest in biological problems and based on a series of three public lectures that he presented to the Irish intelligentsia in 1943 (as one of his statutory duties as the founding director of the Dublin Institute of Advanced Studies), "What is Life?" is a classic example of his exceptional expository skill---in a second language, no less---whereas Baum's book would have profited from another round of copy-editing. But the most striking difference between these two titles lies in the cogency of their respective contents.

Although Max Delbrück and his colleagues had used measurements of mutation rates of fruit flies under X-radiation to show that their genes were necessarily of molecular dimensions in the mid-1930s, the implications of these data were unnoticed by the literate world of the mid-1940s. Thus Schrödinger's public lectures were newsworthy, being favorably noted by Time magazine in the spring of 1943, and his subsequent book---after some difficulties with an Irish publisher and the Roman Catholic Church over the religious implications of his ideas---went on to sell over 100,000 copies for Cambridge University Press, with translations into seven languages. Is there a similar communications gap in our current understanding of the nature of thought?

Noting his background in computer science, one mightclassify Eric Baum among those who believe that ``our souls are software'', but this is not quite fair. Although he states that ``the obvious inability of present-day computer science to account for [the brain's behavior] is no reason at all for doubting that they can be accounted for by computer science,'' the intellectual perspectives of "What is Thought?" are broader than this assertion seems to suggest. The book begins with several interesting chapters on the nature of computation (I particularly liked the presentation of the traveling-salesman problem), which include discussions of the importance of making decisions at the level of semantics, the Turing test, properties of neural nets, hill climbing in a fitness landscape, among several other relevant topics. These discussions lead into the author's central thesis that the mind, like all efficient computer programs, is necessarily modular. In other words, each aspect of the brain's dynamics comprises several subroutines, which presumably can be further broken down into hierarchical structures of nested activities, and he discusses several permutations of this important concept. Curiously, Baum's otherwise comprehensive list of references does not include Donald Hebb's seminal and classic work, in which the notion of ``cell assemblies'' (which are dynamically self-sufficient modules of neurons) was first suggested over a half-century ago. As a psychologist, Hebb aimed to ``bridge the long gap between the facts of psychology and those of neurology,'' and coming at about the same time as the development of the digital computer, his formulation has provided the basis for many numerical studies starting in the 1950s and continuing to the present day which are in accord with a growing body of electrophysiological data. Setting this quibble aside, Baum offers compelling psychological evidence for the modular structure of mind and provides his readers with an interesting and informative account of how the structure of our thinking may have developed over the course of biological evolution, with particular attention paid to computational constraints on the development of learning mechanisms. Importantly, his perspectives are broader than those of many of his colleagues, as he asserts that the ``whole program'' of a brain's dynamics includes the ``complex society'' in which it is embedded. Indeed, the author's evident humility in the face of awesome intricacy of mental activity is, to me, one of the more appealing aspects of "What is Thought?"

The often suggested possibilities for quantum computation are discussed in some detail, along with an analysis of the widely noted example of ``Schrödinger's cat'' which was originally proposed to emphasize the difficulties of applying ideas developed for atomic dynamics to complex macroscopic systems. Considering that a quantum computer---if it is at all possible to construct one---must be carefully isolated from structural irregularities and operated near absolute zero of temperature, Baum joins the majority of physical scientists in concluding that it is ``highly unlikely that quantum computation is relevant to the mind.''

Eric Baum has a dog, and---like most of us dog owners---he is convinced that his pet is conscious, but he goes on to assert that ``we do not need to posit new qualitative modes of thinking to explain human advance over animals. To my mind, the difference between human intelligence and animal intelligence is straightforwardly explainable by cumulative progress once there is the ability to communicate programs.'' Here, again, Baum could profit from reading Hebb's book, which contains but a single mathematical expression, namely A/S. This parameter represents the ratio of the associative area (A) of a mammalian neocortex to its sensory area (S), and it becomes greater as one progresses from rats through dogs to humans. A related physiological parameter---with profound significance for the ease and rate at which modules (or cell assemblies) can switch on and off---is the percentage of inhibitory intercortical neurons, varying as follows: rabbit (31%), cat (35%), monkey (45%), human (75%) [6]. Of course, these relative differences may be examples of the ``cumulative progress'' to which Baum refers.

In a penultimate section, Baum discusses the question of free will, noting that ``our decisions look, from any reasonable perspective short of knowing the exact state of our brains and simulating them in detail, like they are introducing genuinely new information.'' In reaching this conclusion, he may be confused by the continuing tendency of many scientists to overlook a phenomenon called ``sensitive dependence on initial conditions'' first studied by the eminent French mathematician Henri Poincaré and widely observed nowadays by those who study nonlinear dynamic phenomena (chaos theory). As Poincare` famously put it over a century ago:

"If we knew exactly the laws of nature and the situation of the universe at the initial moment, we could predict exactly the situation of that same universe at a succeeding moment, but even if it were the case that the natural laws had no longer any secret for us, we could still only know the initial situation approximately. If that enabled us to predict the succeeding situation with the same approximation, that is all we require, and we should say that the phenomenon had been predicted, that it is governed by laws. But it is not always so; it may happen that small differences in the initial conditions produce very great ones in the final phenomena. A small error in the former will produce an enormous error in the latter. Prediction becomes impossible, and we have the fortuitous phenomenon."

For an author who bases many of his conclusions on close mathematical reasoning and offers a theory that purports to be ``capable of explaining everything,'' the implications of these ``fortuitious phenomena'' should be carefully digested.

Alwyn Scott
http://personal.riverusers.com/~rover/

The main thesis of this book, asserted repetitively, is that the mind is a computer program. Once this is borne in mind, pardon the alliteration, most of the book is reduced to an argument in its favour, rather than an investigation into its credibility. The book often reaches for blunt assertions to support its positions and only afterwards begins a slight retracing of steps. For example, we are told that inductive bias and learning algorithms are coded into the genome. It is obvious, bit of speculation on DNA, evolution and algorithms and out comes the result!

In his observance of Occam's Razor, the author confuses the appeal of the simplest explanatory hypothesis with the belief that he has found such. The discussion of neural networks leaves aside recurrent networks, which are probably more biologically plausible than competitors.

Likewise the idea that the brain essentially 'runs' compressed programs due to evolutionary endowments is unconvincing and philosophically leaky.

I don't want to be over critical of the book as it has brought together many interesting strands of work, but it just has not woven them into anything interesting. There is little new here, whether from modularity or evolutionary programming constraints on neural activity. A lot of it is speculative and several of the key themes are discordant due to under analysis of their assumptions.

Several of the elaborations verge on the frivolous. For example, there is a particularly woolly argument linking the learning of Scheme to "what goes on in constructing our understanding of the world" (p. 222). Likewsie in discussing awareness and consciousness, the author relies on the use of 'main' in C to metaphorically explain how information might come together in the brain (p. 413-415). All kinds of reification fallacies come to mind, leaving aside the thinnes of the argument.

The bottom line is that the book pursues a strong cognitivist program (the brain is a computer) without convincingly examining various sides of the argument. I was certainly no wiser off at the end of it.

In many respects Baum's book is orthodox cognitive science: "the
discussion in this book follows what I perceive to be folk wisdom
among computer scientists interested in cognition." (page 2) In
fact, it is probably the best such text that I've read in years.
I highly recommend this book to anyone studying cognitive systems.

Baum basically agrees with Werbos' definition of an intelligence:
"a system to handle all of the calculations from crude inputs
through to overt actions in an adaptive way so as to maximize
some measure of performance over time" (P. J. Werbos, IEEE Trans.
Systems, Man, and Cybernetics, 1987, pg 7). Or, in Baum's words:
"I am proposing to think about creatures...that are given a reward
function...learning and computing algorithms...The creatures then
apply these algorithms to maximize reward during life." (page 396)
Of course programs that do exactly that have been around for a long
time: "Adaptive systems using learning matrices" (K. Steinbuch and
E. Schmitt, Biocybernetics in Avionics, Gordon and Breach, 1967, pg
751).

In his book Baum frequently equates reward/fitness/utility, U, with
number of offspring a creature has, N. In fact, a more biologically
accurate model (for mammals) might be U=(N-2)/L where L is the
creature's lifespan.

But Baum is quite UNorthodox in that he believes in an extreme
dependence on innateness. He believes that via our DNA we receive
a large number of computational subroutines which contain a great
deal of knowledge about the world.

Baum believes that "semantics comes from compression...If one
compresses enough data into a small representation, the
representation captures real semantics, real meaning about the
world." (page 102) But, unfortunately, a number of DIFFERENT
models may fit the data. As Baum himself admits: "there are likely
many possible locally optimal solutions as good as the one evolution
has come up with that may differ considerably in detail." "There
may be many compact discriptions ...aliens might think of the world
using a substantially different description..." (page 212) So
something which has "meaning" for you, with your model of the world,
may have NO meaning for someone else (having some different world
view). Baum seems to admit as much on page 226: "...there is some
evidence for an evolved module for religious faith, which might well
exist whether or not there is in actuality an anthropomorphic god."
Unfortunately, then all meaning is purely RELATIVE and it makes no
sense for Baum to talk about some "concept really present in the
world." (page 162) Rather, concepts are defined (INVENTED) by
people in the course of their efforts to organize their observations
of the world. Our concepts need not really exist IN the world. They
are best regarded as mental fictions.

Although Baum frequently distinguishes animal intelligence from human
level intelligence he makes no room for the existance of an artificial
intelligence which is not isomorphic to human reasoning. In actual
fact there are many important applications waiting for an artificial
intelligence even IF it were not fully on a par with human reason.
Furthermore, with regard to human level AI Baum seems only to
recognize the ways in which humans outperform computers. Alongside
the list of things people do better than computers one should place a similar list of the many things that computers do better than humans:
computers have better memory, are better at logic, statistics, and
math, can be diskcopied, etc., etc. "What people can't do" (comp.ai,
21 May 1997, R. Jones) I would point out that my Asa H system
(Trans. Kansas Academy of Science, 2006, vol 109, no 3/4, pg 159)
has most of the functionality Baum requires of an intelligence.
It compresses what it learns, is guided by a value function module,
and is hierarchically (self)organized. Perhaps only the vast store
of innate categories is missing; waiting to be learned.

Baum's book is always stimulating and in some ways admirable, especially in its instance that there is nothing magical in the brain. But he's wrong in several crucial ways, the same ways that Pinker get's wrong (for example, in "The Slate's Last Stand").
1. Despite his neural network background, Baum fatally underestimates the power of unsupervised learning. While he's right that complex networks cannot be explicitly trained without astronomically numerous examples, it's now clear that unsupervised learning (where the number of examples is quite literally astronomical) combined with the rather regular (albeit complex) structure of the world, can do most of the heavy lifting, with supervision filling in details. Explaining unsupervised learning to a lay audience is not easy (I know of no successful attempts) but cannot be shirked.
2. Because of his background, Baum fatally overestimates the power of Darwinian evolution. For example, he completely omits the Eigen error threshold problem, he does not take seriously the gap between the information content of genomes and brains, and he seems to think that adding one bit per generation (which is all evolution can do) is a powerful learning procedure.
3. He's hopelessly starry-eyed about the ability of Darwinian evolution to find "compressed descriptions" (though he's spot on in his emphasis on compression). Both evolution and learning are algorithms for adapting, and Baum completely overlooks the possibility that brains can implement the Darwinian algorithm in a different physical medium (synapses instead of nucleotides). To validly draw the conclusions he jumps to, he would have to prove that either the Darwinian algorithm cannot be implemented neurally, or that it would be far too slow (while the evidence suggests that the basic update can be done neurally a billion times faster neurally than genetically). As Dawkins has emphasised, Darwinism is the only way to get intelligence, but this does NOT mean that only DNA can do it.
In sum, a book for the beach, not for eternity.

This book covers data mining at a serious level, including essential material on testing and a broad array of techniques. It is written for practitioners and provides clear explanation of included topics. Easily one of the best 5 books on data mining currently available.

Note that this book has moved on to a second edition.

I have bought this book as course book to learn some particular aspects of data mining.
With the software that you can dowload you can do yourself all the exercices for every models presented
It's the best way to progress
Do the same, it's simple and funny
The explanations are very clear and pedagogical, very practical

This book is actual a textbook for a data analysis course. We use it because the flow of the chapters is almost the same as the flow of the course material. Unfortunately, it is not as useful as expected if you are in the field. It is not in depth for the materials that the authors wanted to cover due to the fact that this is not a book for just programming or just statistics. If you have a strong background on machine language or a strong background on data analysis, you may not find it useful for you career. This book is for those who have limited knowledge on both programming and statistics, but not for professionals.

Witten's book, combined with the accompanying open source package, Weka, provides a great overview of data mining principles and practice from a machine learning perspective. One could hardly do better than to own this book and "The Elements of Statistical Learning; Data Mining, Inference, and Prediction" by Hastie, Tibshirani, and Friedman which covers complimentary material from the statistician's perspective. Witten does an amazing job of providing a comprehensive overview of the field while still providing some depth re. the algorithms; after reading the book I didn't feel like I'd read yet another large volume of empty claims about the power of information technology to make me rich and famous. In fact, with the book by my side, in a relatively short time I was able to use Weka to pry some useful information from one of my medical imaging data sets (maybe even enough to serve as preliminary data for a grant application). It seems to me that with an understanding of the material in this book and the one by Hastie et. al. one could embark on serious data mining projects.

I've been working with "big name software" for some years, but when I joined the institution I work now and no tools where available I begun my quest for an open source tool that could help me build statistical models applied to real business problems.

As a result of this quest I found the WEKA data mining software on the Internet (you can find it on www.cs.waikato.ac.nz/~ml/weka/) and that nice piece of software leaded me to this book.

This book is EXCELLENT and I am giving 5 *five* stars to it as it helped me understanding the whole process of datamining: from loading the data to building the model.

I've read some reviews and I think some of them are not fair (particularly one that says that this book have "just words with no relation or sense at all").. THIS BOOK IS REALLY WELL WRITTEN but you have to read it slowly: As when you study something.

Buy this book (*don't forget to download the software*) and I am totally sure that you will be producing and using models in a week.

Can't imagine that some weeks ago

Cheers,

I have to agree with all of johnnied7 criticisms. This book is pitched at a level too advanced for an introduction. It also reads and is structured like a research paper. Not recommended.

I have an engineering degree, and I found this to be a little tough to follow for two reasons:

1. Not enough step by step prodecure especially at the beginning. Mitchell is too quick to start with the math formulas. It turns out that Genetic Algorithms are fairly straight forward and easy to follow, but you have to read this book twice before you "get it" because Mitchell clouds the discussion with proofs and mathematical representations of systems. It is tough to follow.

2. Mitchell does a poor job of selecting meaningful examples to illustrate the points. A nice simple set of examples where the average person easily picture the system would have been delightful. Instead this author chooses to illustrate the Genetic Algorithms through uncommon neural networks amoung other exotic applications. I found myself struggling to understand both the example (I didn't know a thing about neural networks!) and the genetic algorithm.

When buying an Introduction type book, I expected it to be more 'down to earth'. this book is for advanced minds!

This book primarily deals with the theoretical side of genetic algorithms. If you are looking for practical knowledge of how to implement a GA you should look elsewhere. For all intents and purposes this is a textbook. It's heavy on theory and proofs, but doesn't always explain everything in depth (that's what class time is for). There are problems at the end of each chapter that can be assigned to students.

There are case studies of many academic projects that seem to drone on forever and aren't really that useful in helping you learn how to write your own GA. Chapter 1 gives an overview and provides all of the appropriate terminology. Chapter 5 gives an high-level overview of how to implement a GA. Those are the 2 must-read chapters, all of the others can be used as torture for CS students.

To recap, if you're teaching a class in artificial intelligence this book is good. If you're trying to figure out how to implement a GA to solve a practical problem not so good. That evens out to 3 stars for my rating. I recommend searching the web, there are a few good sites on GA programming.

First it must be said that the book is not an introduction that the non-scientist will easily understand. Some knowledge of computer programming is assumed. It acknowledges this in the last paragraph of the preface. Many of the notations in the book are unfamiliar to business or financial readers. There is no mathematics beyond algebra so the aforementioned prerequisites are the main hills to climb.

Mitchell's book is an overview of genetic algorithm analysis techniques as of 1996. The author gives a history of pre-computer evolutionary strategies and a summary of John Holland's pioneering work. A description of the basic terminology is presented and examples of problems solved using a GA (such as the prisoner's dilemma). The second chapter discusses evolving programs in Lisp and cellular automata. Also included in this chapter is a discussion of predicting dynamical systems. This was the section that has the most interest for me. Also interesting was the summary in this chapter about putting GAs into a neural network so that the ANNs could evolve.

The fifth chapter discusses when to employ a GA for maximum success. I appreciate the clearly thought out discussion of when to choose a GA for a problem. Sometimes authors of these types of books mimic the man with a hammer that thinks everything looks like a nail.

This is a great place to start to learn about genetic algorithms. The writing is clear and not bogged down by jargon. The book is not overly technical; it is written for the layman and has a casual conversational style that is a pleasure to read.

About half of the book is devoted to presenting examples of studies that have used genetic algorithms. These examples are interesting in themselves and also serve to illustrate the variety of genetic approaches that are available. The book also presents conflicting points of view of experts about which algorithms work best and why. This is helpful in combatting the impression that a beginner sometimes gets that everything is simple and all the answers are known.