Memory and the Computational Brain: Why Cognitive Science will Transform Neuroscience 1st Edition

There are a very small number of books that made a major change in the way I think. Chomsky’s Syntactic Structures changed the way I think about language and philosophy. Kent Beck’s Extreme Programming Explained changed the way I develop software. Dawkins’ Selfish Gene gave me my first true understanding of the theory of evolution by natural selection. Memory and the Computational Brain is one of those books. It changed the way I think about psychology.The essence of the book is that the proper model for human cognition is not neural networks but a Turing Machine. A Neural Network is a (very complex) Finite State Machine (FSA). It is inherently more limited computationally than a Turing Machine because it can't represent historical data. This is the reason that Turing created the Turing Machine model for his paper on the Entscheidungsproblem (On Computable Numbers), he realized he needed a formalism more powerful than an FSA that included memory.The question that Gallistel and King raise is long term memory. Not muscle memory which is described by Hebbian conditioning (changing the conductance and connections of a neural network based on learning) but episodic memory such as “who is the current president?” The authors make a strong case that there is no good neural net model for this type of memory and that the structure of such networks makes them unsuited to representing historical data in a scalable manner. Of course, with computers this issue never arises because Artificial Neural Nets (ANNs) just use the computer memory (e.g., arrays, databases, spreadsheets). However, many connectionists ignore this issue when it comes to the brain. The authors provide detailed analysis of the proposed neural network based solutions to episodic memory and show that such approaches could not scale to the memory requirements of Scrub Jays (which are known to cache food for the winter in tens of thousands of locations) let alone humans.When one gives this a bit of thought it seems (at least to me) very intuitive. Of course, the output of one network can be the input for another but ultimately the input and output of all ANNs is some form of data which is virtually never represented as a neural network and it seems intuitive that the brain must function in a similar way, i.e., have a different mechanism for storing the input and output of neural nets that is analogous to addressable computer memory. The authors provide strong arguments for this view.They hypothesize that the mechanism for storing episodic memory is some form of molecular code such as RNA or DNA. The advantage of such a code is that it is orders of magnitude more efficient than an FSA for storing data. Such a code would help to address one of the biggest issues with computing by neurons: computation by neurons is orders of magnitude slower than computation with computers. E.g., the interval between "spikes" (signals from other neurons) to the action potential (firing) of a neuron is on the order of half a millisecond and the time for a neuron to return to its base state after firing takes several milliseconds. A modern computer can execute at least one floating point instruction in .001 of a microsecond (a microsecond of course = .001 milliseconds). Yet the brain of a child can outperform computers on many tasks.Anyone interested in psychology, whether cognitive, evolutionary, neuropsychology, or behaviorism should read this book. Even if you completely disagree with the authors, I think you will find the questions they raise fascinating and their elucidation of concepts from computer science such as Shannon’s Information Theory and the theory of computation extremely clear as is their description of how such concepts are essential to a theory of human cognition.

The argument of this book is water-tight:1. The brain is a computational device2. All efficient computational devices have a processor and a tape (a read/write memory)3. The brain is efficientConclusion: The brain has a processor and a tape (a read/write memory)No (computational) neuroscientist would disagree with 1. No computer scientist would disagree with 2. But because neuroscientists know (and this is true) that synapses are noisy and inefficient, and believe (as a matter of Humean faith) that the synapse is the only candidate for memory storage in the brain, they have to deny 3. As a result, neuroscience rejects the conclusion. That's why Gallistel and King spend so much time considering brains like those of scrub jays, bees and ants - all of whom perform memory feats that mathematically belie the idea that reverberatory synaptic loops are the only way to encode memory. There's even a section that persuasively argues that Pavlov's dog is NOT associating (in the sense of Hebbian rewiring). An incredible book that will change your mind if you are not part of the church of Hume.

Gallistel correctly argues that synapses are too inefficient to act as the "Turing tape" that is necessary for (symbolic) computation, though his reasoning is wrong: the real problem with synapses is that their plasticity interacts, as a result of their extremely close-packing (which is precisely what makes them potentially so useful). This "crosstalk" can undermine sophisticated, quasi-symbolic, synaptic learning. But his proposed "solution", that some unknown new neural storage process analogous to DNA underpins powerful quasi-symbolic brain computations, is pie-in the-sky. Nature is a tinkerer, and it seems much more likely that she has simply patched up the unavoidable defects of synapses using largely ready-made materials. In particular, it's likely, though not proven, that the neocortex is specialised to implement a type of "synaptic proofreading", which allows synapses to act as symbols (see syndar dot org). And the same basic idea, proofreading, also underlies the extraordinarily accurate copying process that underpins Darwinian evolution. So "mind" would be a synaptic version of "life".

The authors' thesis is that since the brain computes, it has to instantiate known information processing and computational principles. They discuss many of these principles and crituque current neuroscience computational assumptions and models as paying insufficient attention to them. They present several biological cases in some detail to illustrate how the current connectionist/neuronet/synaptic weight approach is insufficient to the challenges these cases present. Finally, they they describe specific functions and processes we should be looking for in the brain to explain the behaviors presented in the cases.It is an excellent book and in the end may warrant its subtitle, "Why Congnitive Science Will Transform Neuroscience." I highly recommend it to anyone who works in neuroscience and especially those who work in computational neuroscience.

The authors clearly present iron-clad computational arguments in a clear and approachable way. This book is wonderful for anyone interested in how the mind works with the willingness to entertain computational arguments (even without a strong background in mathematics, though that helps).In the end, this book underestimates its impact: if the mind essentially processes information and we do not even understand the way that information is written or stored, how can we make the grand claims to knowledge of the mind one sees all over popular science today? How much ink is everyone wasting on grandiose claims when they cannot even explain the basics?