1 of 1 people found the following review helpful:
4.0 out of 5 stars
Things have changed a lot since this book appeared., February 21, 2004
This review is from: Artificial Intelligence (Handbook Of Perception And Cognition) (Hardcover)
This book is a collection of articles that give a fair representation of the status of artificial intelligence in the mid 1990's. In only a decade since those times, the field has expanded considerably, mostly due to applications and the rise of the Internet. Controversy as to the nature and characterization of machine intelligence continues of course, and one finds both intense criticism and uncritical optimism of the nature and future of artificial intelligence. Most of this takes place in the philosophical literature, but there are also places in mainstream AI circles where predictions of future developments are overly optimistic. This optimism is refreshing but it can be a distraction for those that are seriously working to develop useful applications of artificial intelligence. But also, cynicism and negativism have also thwarted research in AI, there having been a few cases that showed much promise, but were abandoned because the researchers were convinced by others that their ideas were unsound. The AM and EURISKO efforts in automated mathematics, which are discussed briefly in this book, are good examples of this.
All of the articles in this book are interesting, but there are a few that stand out due to their penetrating insight on matters that are still of great interest in artificial intelligence. One of these is the article entitled "Creativity" by Margaret A. Boden, who is the editor of the book, and who has done some outstanding work in the elucidation of what it means for a machine (human or otherwise) to be creative. Her insights on this subject are many, and in the opinion of this reviewer her works should be required reading for all those interested in the origins of creativity and attempts to implement it in non-human machines.
Boden asserts that every case of creativity cannot be explained by a single scientific theory, one reason for this being that an AI model must be evaluated, and such an evaluation is outside the realm of science. In addition, there is a high variability in creative psychological processes, which preclude a general understanding of them. Lastly, creativity is very idiosyncratic but Boden is careful to point out that it is not random, but instead subject to constraints in `conceptual space'. Boden does not define conceptual spaces from a mathematical standpoint in her article, but she does discuss their utility in modeling creativity in AI, and the role of AI models in making more rigorous the conceptual spaces employed by musicologists, literary critics, etc. Boden's ideas in this article on the mapping, exploring, and transforming conceptual spaces can be viewed in the context of dynamical systems, and such a view allows them to be more easily coded into a machine language. Boden discusses several examples of AI models of the arts, such as connectionist models of music, the AARON program for generating line drawings, and the Letter Spirit project, which tries to model the perception of alphabetic style. She points out the pluses and minuses in each of these examples, such as the limited compositional ability of connectionist models, the limited evaluative and self-correcting powers of AARON, and the lack of Letter Spirit in being able to justify its own decisions. AI models of science are also discussed, and Boden concludes that most of these are `data driven' and cannot identify relevance for themselves. She believes that they can learn, but their discoveries are `exploratory', and do not succeed in changing their own conceptual spaces.
The status of machine intelligence for scientific discovery has changed quite a bit since this article was written. Techniques from inductive logic programming coupled with faster hardware are making the reality of automated scientific discovery closer with every passing year in the twenty-first century. With more powerful hardware on the horizon, these developments give evidence of a time when drawing on the efforts of their human tutors, machines will be able to think across scientific domains and formulate hypotheses and creative ideas that may far surpass anything that has been done by human scientists. The time scales needed for this scientific discovery to take place may be so short that it might be difficult for human observers to assimilate these new results in order to evaluate their efficacy and applicability. The machines though may have their own opinions on the utility of the ideas and theories they derive.
Help other customers find the most helpful reviews
Was this review helpful to you? Yes
No
