This book crosses disciplinary boundaries by melding ideas from the biological study of evolution with the notion of systems thinking in view of formulating a perspective on western culture's understanding of complex systems and their emergent behavior. Kelly truly believes, and provides evidence for, the fact that our machines are becoming more biological in nature, while our natural biology is becoming more engineered. In essence he argues that our world is metamorphosing into large living entities, or systems, where boundaries between nature and machine are more of an intellectual ideality than a physical truth.
The idea of complexity in this book is addressed from the point of view of collections of elements that interact based on a set of basic operational principles, but whose collective behavior is an emergent property that cannot be inferred from the behavioral properties of the basic subunits of the system. Unlike a hologram, which when cut into n pieces produces n holograms of the same image, the complex system, or collective, exhibits a behavior that is completely different from that of its elements. Kelly claims that all of these complex systems are of a distributed nature, and have the advantages of being adaptable, evolvable, resilient, and boundless, yet suffer from apparent non-optimality, uncontrollability, unpredictability, and appear incomprehensible.
Throughout the book, Kelly traces historical references to the question of emergent behavior in natural systems. Much of the discussion focuses on biological systems, such as bee hives, swarms, and ecosystems of varying scales. While there is no central control authority associated with any of these systems, they are able to demonstrate "collective thinking", where the system will adjust and change in response to external stimuli. In fact, it is argued that it is very easy for collections of natural elements to evolve into a semi-permanent equilibrium state that exhibits behavioral characteristics of a new entity. However, the true, if insurmountable difficulty is in combining the system elements in such a manner that the system exhibits a desired behavior or set of behavioral characteristics.
In parallel to the examples from natural systems, Kelly focuses on many examples of complexity from the world of computing and network communications. While traditionally humans have taken physical goods from nature for survival, Kelly believes that the next step in understanding and creating complex systems will arise from human ability to capture, replicate, modify, and employ the logic associated with natural systems. Kelly recounts many examples of robots, software, or computer-human systems that scientists have used to emulate the behavior of natural systems. The reproduction of evolutionary changes apparent in natural systems via the use of distributed, modular software programs is a proof that Kelly retrieves in many examples. The ability of a set of basic computer programs to replicate the complex population lifecycle behavior of interacting species in a closed ecosystem, where each program is only designed to perform a minimal set of survival functions, helps open the reader's eyes to emergent behavior in a complex system. The next question this leads to in the book is the issue of controlability.
The history of introducing control into machines begins in this book with Ktesibios' water clock flow governor, and progresses through thermostats, steam engine regulators, and Weiner's idea of Cybernetics with humans in the loop control. The basic tenet of the discussion of control was that systems that have tightly coupled variables, and have one that is controllable, also provide a good opportunity for creating a control mechanism for the system. This may be fine for very linear, local, closed systems, but as the systems become more distributed and complex, the number of variables and apparent connectivity become impossible to manage. Kelly argues that complex systems can only be managed or controlled by relaxing any form of central authority or imposed control mechanism. This is where the title of the book comes from, "Out of Control" where Kelly quotes Lao Tzu in saying that "most intelligent control methods appear as uncontrolled methods".
However, the only guide that is provided on how to control these systems is that they must be built slowly from their parts, ensuring functionality and predictability at each level of system integration, until a meta-system is assembled. This is reflected Kelly's discussion of Rod Brooks's methodology for designing complex systems (robots in his case), by using the following steps:
Do the simple things first
Learn to do them flawlessly
Add new layers of activity over the layers of simple tasks
Make the new layer work as flawlessly as the simple layer
Repeat ad infinitum
Yet, Kelly then spends several chapters discussing network economics, electronic commerce, and the predictability of complex systems and essentially says that no matter what you do, once you build a complex system, or one establishes itself naturally, it is impossible to understand the determining factors of the emergent behavior. At best you can observe short-term trends, but the system will operate as an independent entity.
Returning to his theme of the rise of neo-biological systems, Kelly provides a review of some of the post-Darwinism thoughts on the emergence of species and the associated natural and computational proof that apparently contradicts Darwin's theories. In essence, this discussion focuses on the fact that the planet has not been populated by individual, independent species evolving over time, but rather interdependent elements of systems that change in response to environmental stimulus, and, in turn, cause environmental change as a result of their adaptation. He truly believes that the interaction of machine, science, engineering, and nature is an interdependent set of action-reactions that are ultimately part of a single living neo-biological ecosystem we know as the Earth. Our actions shape the system and the system shapes our actions.
While a very intriguing book with many interesting vignettes, I found the discussion was more focussed on complexity and systems, and less on the evolution of nature-machine as the first half of the book promotes. The issue of control in complex systems is a very important aspect of this field of study, yet there is very little concrete evidence or methodology provided for managing complex systems, other than the fact that Kelly keeps saying that central command control does not work. From an entertainment standpoint the book is an easy read and offers many stories to excite the reader. However, from a scientific perspective, the book is more a collection of "stuff about systems, complexity, and evolution" than a succinct treatment of the controlability of complex systems.