The Age of Intelligent Machines is a non-fiction book about artificial intelligence by inventor and futurist Ray Kurzweil. This was his first book and the Association of American Publishers named it the Most Outstanding Computer Science Book of 1990. It was reviewed in The New York Times and The Christian Science Monitor. The format is a combination of monograph and anthology with contributed essays by artificial intelligence experts such as Daniel Dennett, Douglas Hofstadter, and Marvin Minsky. Kurzweil surveys the philosophical, mathematical and technological roots of artificial intelligence, starting with the assumption that a sufficiently advanced computer program could exhibit human-level intelligence. Kurzweil argues the creation of humans through evolution suggests that humans should be able to build something more intelligent than themselves. He believes pattern recognition, as demonstrated by vision, and knowledge representation, as seen in language, are two key components of intelligence. Kurzweil details how quickly computers are advancing in each domain. Driven by the exponential improvements in computer power, Kurzweil believes artificial intelligence will be possible and then commonplace. He explains how it will impact all areas of people's lives, including work, education, medicine, and warfare. As computers acquire human level faculties Kurzweil says people will be challenged to figure out what it really means to be human.
Background Ray Kurzweil is an inventor and serial entrepreneur. In 1990 when this book was published he had already started three companies: Kurzweil Computer Products, Kurzweil Music Systems, and Kurzweil Applied Intelligence. The companies developed and sold reading machines for the blind, music synthesizers, and speech recognition software respectively. Optical character recognition, which he used in the reading machine, and speech recognition are both featured centrally in the book as examples of pattern recognition problems. After the publication of The Age of Intelligent Machines he expanded on its ideas with two follow-on books: The Age of Spiritual Machines and the best selling The Singularity is Near.
Content
Definition and history
Kurzweil starts by trying to define artificial intelligence. He leans towards Marvin Minsky's "moving frontier" formulation: "the study of computer problems which have not yet been solved". Then he struggles with defining intelligence itself and concludes "there appears to be no simple definition of intelligence that is satisfactory to most observers". That leads to a discussion about whether evolution, the process, could be considered intelligent. Kurzweil concludes that evolution is intelligent, but with an IQ only "infinitesimally greater than zero". He penalizes evolution for the extremely long time it takes to create its designs. The human brain operates much more quickly, evidenced by the rate of progress in the last few thousand years, so the brain is more intelligent than its creator. Kurzweil concludes from this that there is no theoretical reason why the human brain cannot create something more intelligent than itself, and suggests "a sufficient number of decades or centuries into the future" humans will in fact be surpassed by their creations. The field of artificial intelligence presupposes that the human brain is a machine, and that an alternative implementation could be built, as a computer program, which would have the same faculties as the real thing. Kurzweil traces the philosophical underpinnings of this tenet, as well as the opposing view that properties such as consciousness and free will are unique to the human mind. Kurzweil starts with Plato and touches quickly on Descartes, Newton, Kant, Wittgenstein and ends with Hubert Dreyfus. Kurzweil also presents the mathematical roots of artificial intelligence including contributions by Bertrand Russell, Alan Turing, Alonzo Church, and Kurt Gödel. The Turing test is introduced as a way to gauge whether the field of artificial intelligence has succeeded or not.
Chess and pattern recognition
Kurzweil discusses how computers play chess in detail, building to his prediction that "we will see a [computer] world champion by the year 2000". The Chinese strategy game of go, however, has proven much more difficult for computers to play well. He considers go to be a "level 3" problem, the type of problem where there is no single unifying formula which solves it. Then Kurzweil reveals that pattern recognition, which is crucial to artificial intelligence, is also a level 3 problem. Kurzweil traces various ways of doing pattern recognition, from the rise and fall of perceptrons to random neural nets and decision trees, finally explaining that intelligence is a hierarchy of heterogeneous processes "communicating and influencing each other". He believes Marvin Minsky's Society of Mind and Jerome Lettvin's society of neurons are useful models. Kurzweil differentiates logical thinking from pattern recognition, and explains that AI has had much more trouble with pattern recognition, exemplified by efforts to create artificial vision. Kurzweil estimates that the human vision system does the equivalent of 100 trillion multiplications per second where "a typical personal computer" of the day could only do 100,000 multiplications per second. The way out of this dilemma is parallel processing, having millions or billions of simultaneous processes all computing at the same time, something Kurzweil felt would happen in the future. Kurzweil also discusses speech recognition, which like vision requires complex pattern recognition.
Knowledge and art
… excerpt ends here. Continue reading the full article.
