POINT OF VIEW
My standpoint as a reader of the book I would describe as 'experienced in algorithms', an expert in learning systems, but an amateur in music production - including computer-generated - and composition. My special interest in reading the book was the question of which methods are currently used to enable computer-generated music. Specifically I'm interested in how a computer program may be made able to 'learn' to play and and to compose music.
A GREAT WEALTH OF MATERIAL
The book 'Algorithmic Composition' provides undoubtedly a very broad introduction to the topic. The concepts of algorithm, markov model, generative grammar, transition networks, chaotic systems, Lindenmayer systems, genetic algorithms, cellular automata, artificial neural networks, and artificial intelligence are discussed conceptually, and examples are given for each chapter. At the end of each chapter you will find final assessment, where the strengths and weaknesses of the discussed algorithms for the purposes of the composition are pointed out.
ALGORITHMS AND COMPOSITION
Perhaps it is no coincidence that the book does not begin with a detailed representation of the compositional process, but with a very extensive chapter (almost 60 pages) about the development of the human sign systems, the concept of algorithm, the development of machines for computation (computers), and the foundational issues of computability and decidability by a computer. What is missing is a similar introductory chapter about to the composition process. While the introduction about the concept of an algorithm provides a good framework for all the subsequent chapters, the concept of 'composition' remains a little 'in the air'. Even in the final discussion this gap is not really closed. Possibly does this lack express the limits of a representation of such a complex topic as 'algorithmic composition' , otherwise this can be seen as a challenge for a possible future revised edition, in which the concept of a - certainly idealized - compositional process will be elaborated to an extend which allows a better interaction with the different algorithmic methods.
Within each chapter, as in the concluding discussion, the book repeatedly makes clear that between the great diversity and complexity of challenging pieces of music and the possibilities of algorithmic formalization has always been gaps, which will not be closed easily by a formalization. This is due not only to the limits of formalization - which has not yet been fully exploited - but also due to the conceptualization of the phenomenon of composition as such. We are always able to play or to compose music, but not so readily we are able to talk the same time about our doings in an 'appropriate conceptual' way (even 'great' composers like Bach, Mozart, Beethoven would not have been able to describe their process of composition in each case in a sufficient theoretical manner).
While the preceding chapters are quite clear about the different algorithmic methods, the chapter on 'Artificial Intelligence' is unsatisfying. Nierhaus himself points out at the beginning of the chapter that the very notion of 'intelligent' or 'intelligence' is used very inconsistently in the literature. This short chapter about AI is no remedy from this fuzziness of the field (one should notice that the algorithmic concepts of all preceding chapters are today in use within the context of so-called 'intelligent' systems or in the context of 'machine learning', the preceding chapters therefore deal already 'somehow' with 'artificial intelligence').
Despite the criticisms in detail, I would like to award the book five stars because it deals with a complex issue in a way that gives a good inroad for everybody who wants to understand the field a bit better. In particular I appreciate the thoroughness of the individual chapters combined with helpful further readings.