Surfing Uncertainty: Prediction, Action, and the Embodied Mind

I think the main hypotheses are true, but this book didn't prove it to me, and it's probably quite difficult to get further without massive simulations.

Great concept, but very heavy reading and quite repetitive. The ideas could have been conveyed in about half the actual text.

A nice and very innovative account of the human mind seen as a Bayesian learning machine comprising an endless ladder of hiearchies. Exciting and inspiring read.

Andy Clark is Professor of Logic and Metaphysics in the School of Philosophy, Psychology and Language Sciences at Edinburgh University, as well as author of several books on the workings of the human brain and cognitive science in general (" Being There: Putting Brain, Body, and World Together Again ", " Supersizing the Mind: Embodiment, Action, and Cognitive Extension ", " Mindware: An Introduction to the Philosophy of Cognitive Science ", etc). His latest book, "Surfing Uncertainty: Prediction, Action and the Embodied Mind", expands and updates the ideas contained in his earlier books by reviewing the latest thinking with regard to human cognitive processing that is emerging from studies at the intersection of neuroscience, psychology, philosophy, artificial intelligence and robotics. The book expounds and develops a predictive processing ('PP') model for brain functioning which posits human cognition as the product of a hierarchical Bayesian predictive coding engine, endlessly generating probabilistic generative models matched against context-variable precision-weighted input streams -- contributed by the body's exteroceptive, interoceptive and proprioceptive sensing systems -- in a complex cycle of top-down predictions and precision estimations, modulated by a counter-flow of bottom-up residual errors. In short, this model suggests that we perceive the world by first predicting how it should be, based on a series of prior expectations and understandings, using our senses to test the correctness of this prediction and iteratively refining our prediction as necessary to minimise the errors. Only when our cognitive systems finally report that top-down predictions satisfactorily match bottom-up observations does the brain's latest refined view register as our current perception of the world and our state within it.Central to this model is a view of the brain's role as very different from that traditionally assigned to it by many cognitive scientists. Rather than have the brain as a passive observer, the PP process requires the brain to take on a much more proactive role, not only constantly predicting its most likely sensory input but also driving the body towards the best position to gather that input. The essential feature here is that this positions the brain as a central agency for action and engagement with the world, ensuring that the human animal is best positioned to "surf the uncertainties" of our sensory arrays, not only for survival in a world by avoiding its hazards and exploiting its opportunities, but also for structuring the world itself and altering the very things with which our brains must engage and predict.What emerges in this book is an elegant and stunningly unified (as well as simplifying) vision of the way the body's cognitive system functions, able not only to model with great precision many of the empirically observed characteristics of human cognition and the way our bodies interact with the world so fluently but also to explain very neatly features such as the autistic spectrum and to shed light on mental illnesses such as schizophrenia. It illuminates other facets of the human condition also, such as dreaming, imagination and the creative urge. And it makes perfect sense from an evolutionary perspective, too. It should be stressed that these ideas are not entirely new, nor are they the author's own: PP is a model of cognitive processing that has been gaining traction for some time now -- for an earlier position of the field, see Jakob Hohwy's 2014 volume, " The Predictive Mind ", for instance -- but Andy Clark's survey here is more comprehensive and deeper delving than most others on the subject.Given its weighty, highly theoretical subject matter, this book is not surprisingly a fairly heavy read. Although the author does his best to keep his text pitched as much for a lay readership as for an expert one, the sheer complexity of many of the ideas and notions that he covers necessitates a certain level of reliance on jargon and the employment of precise technical language; the result otherwise would undoubtedly have been an undesirable degree of ambiguity and vagueness that would have served no purpose whatsoever. The book can be read by an interested lay reader but will still require considerable investment of effort if anything is to be gained from it.The serious student of the subject should find within its pages an excellent position statement on the current development of PP theories, together with a cogent and succinct analysis of its current shortcomings (few) and useful directions for future investigative activity. Particularly welcome also are the 75 pages of additional notes, appendix and full references, as well as comprehensive index.

I am a big fan of lay science books and have a modest knowledge of neuroscience and psychology so was looking forward to reading this book. I also have a working understanding of Bayesian methods. I’d say that these are something of a prerequisite to get the most out of this book. At first I was disappointed by the unnecessarily inaccessible language - especially during the introduction (which was most probably written after much of the content of the book). It was not a problem of jargon or terminology as I am reasonably familiar with this subject and the predictive processing concept is not a new one to me; the problem lay in the run-on sentences peppered with overly elaborate terms for concepts that could have been expressed more succinctly. Writing for a lay science readership is something that relatively few researchers master. Fortunately, once the book moved on from the turgid introduction, it began to flow with a much more manageable style, carefully detailing the concepts and supporting evidence with particularly good attention to referencing and a few useful appendices. The arguments in favour of Clark’s viewpoint are persuasive and with a solid book like this to refer to, it has a good chance of taking hold across a wide variety of disciplines from neuroscience to AI to human-factors engineering. As I say, much of what is within has been highlighted to a non-specialist reader before – some elements covered in  Connectome: How the Brain's Wiring Makes Us Who We Are  ; others in  Thinking, Fast and Slow  for example – but the synthesis is thorough and there was plenty that was new to me. Very worthwhile reading and don’t be put off if it seems hard going to begin with.

Andrew Clark’s authoritative book reveals the incredible processes within our “meat computer”, through which the miracle of perception occurs. The idea that we do not have a clear, immediate and direct access to the rich world around us, but somehow actively hallucinate what we experience, is counterintuitive, yet here is the mountain evidence from what we are now learning of the brain’s neurological structures and connections.Andrew Clark builds the rigorous evidence that our brain is a predictive processor, using top-down predictions to correctly guess or "explain away" bottom-up sensory information in an iterative, hierarchical manner, and that differences between what we expect and sense, the "prediction error," travel upward to help refine the accuracy of future predictions. He also describes how through our actions in the environment we direct our attention to minimise the prediction errors, (and maximise the energy efficiency of the process).This book will be vital to academics, being crammed with details, references to the vast range of work in this area. This is no popular science book, yet as someone who has only an understanding of the human mind from an external connectivity perspective, I found it quite fascinating.Richard Epworth, author of “Bottleneck - Our human interface with reality”