How the Body Shapes the Way We Think: A New View of Intelligence Book

It is common feeling that intelligence is one of the last features that seem to stand to mark humans in comparison with lesser entities. The rise of putative machines and robots that hint at overshawoding such statement is then considered, by some, as the next frontier of anthropocentric nihilism, by others as the welcome beginning of the rise of trans-humans (in the jargon of Kurzweil-feeded believers in Singularity). As it is set so far, anyway, this perturbing perspective seem to be grounded in frail foundations - said better, on very strong and yet narrow and possibly mislead conceptions about the nature itself of intelligence, its origin and how it can be emulated, reproduced if not enhanced. Gelernter's inspiring and thoughtful "The muse in the machine" provided some early warning to mainstream artificial intelligence, noting how computation-only processes are bound to just scratch the surface of what intelligence (with particular emphasis on creativity and associativity, in his case) is and can output. Pfeifer and Bongard purport a far deeper-reaching assault on what has come to be known half-mockingly as "good olf-fashioned AI" (GOFAI) - that is, the classic view that intelligence is the sole result of symbol processing based on internal representations of solution spaces.
The authors build their arguments on over 20 years of counterstream research started over by revolutionary MIT's AI grey eminence Rodney Brooks and currently spread across many institutions - particularly Japan where the actual interlacing of human and artificial entities has basically always been culturally welcomed, as opposed to Western cultural biases toward full control by natural products. Brooks spawned this rapidly growing and extremely promising field of research with provoking papers titled "Intelligence without Reason" (brilliant pun intended) and "Intelligence without Representation". The key to the new paradigm is the centrality of embodiment in processes that are normally assumed to be cognitive. The body is, in short, essential to intelligence, and abstract entities are bound not to reach comparable level of performance nor intelligence with as little resources as embodied ones. This perspective was supported by and in turn directed much attention toward the study of social, interacting animals like birds, ants, bees and the likes. It cuts through an old line of thought which has always seen the brain as the exclusive center for high-level cognitive processes and such processes as simply based on symbolic computations, in principle reproducible by machines as abstract as to be even unconnected to the real world. Such idealism is implicit in Turing's landmark works of 1930's which directly inspired the architectures of computers, whose spirit is to a fair extent kept unchanged till today. More importantly, it is deeply rooted in the common understanding of intelligence itself, including memory, creativity, plasticity, evolution and development. Brooks rejected many if not all such assumptions, and started looking from an opposite, more humble and bottom-up perspective by proposing the subsumption architecture for controlling intelligent-looking robots, and forging impressive slogans such as "The world is its own best model".
Pfeifer and Bongard, two of the main players of today in embodiment, develop the implications of such momentous shift in focus by recalling the current landscape of AI and challenging the (implicit and explicit) assumptions by showing how the presence of a material, physics-compliant body is unavoidable and central for all kind of intelligent behavior. They do not (and still can not) propose a solidified and all-encompassing theory, still they provide and exemplify at length many design principles for both individual and collectives of embodied agents, and for short- to long-term time frames. Ecological balance, cheap design, design for emergence, sensory-motor coordination, frame-of-reference, redundancy, multiple parallel and loosely-coupled processes, morphological computation and evolution are notable examples of such principles, which they claim can in fact pertain and give novel insights to all agents given a proper embodiment (including nodes in ubiquitous computing networks, and firms). The principles are introduced in the first two of the four parts of the book, which turn out to be the most interesting and absorbing in hindsight. The third part is in fact devoted to exemplifications (ubiquitous technology, firm management, memory and daily-life robotics) and, with the exception of the chapter on memory perhaps, it indulges in excessive verbosity and re-statements of at-this-point acquired notions. The forth part recapitulates the content of the book pretty nicely and lucidly, without loosing the sense of urge for an enormous amount of new research to be conducted and the excitement for the potential that the embodied perspective can still provide for deeper understanding of what intelligence (and eventually life, both in their biologic and artificial versions) is and how it could otherwise be. (It notably ends with a suspended excerpt from a taoist-inspired novel.)
The book is very well documented and referenced. We only feel puzzled at the remarkable absence of mention of the powerful "analysis by synthesis" methodology developed by the group of Lipson (who is still nonetheless cited several times for other works in the book) also before the publication of the book (with particular reference to Zykov's co-evolution of the control and of the representation of the body by an actual four-legged robot). It is also illustrated by several cartoons and pictures, and features boxes for additional in-depth discussion of selected topics. Though much focus is devoted to Pfeifer's own AI lab at ETH Zurich, groundbreaking works from most active groups are precisely mentioned and illustrated (e.g. walking robots, delayed reward learning with artificial mouse in T mazes, entertaining and "healing" robots, robots with facial expressions, sort of chimeras).
The book addresses curious laymen with a hunch for provocative thought and counterfactuals. It's a book against the grain written in a plain though at times too redundant style - yes, it could be made a bit shorter. It inscribes to that extremely valuable set of books that informedly point outside the mainstream for more light on dark issues. Very recommended.