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 the 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 [...]). 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".

Читать «Memory and the Computational Brain: Why Cognitive Science will Transform Neuroscience»

Купить электронную версию