In a tragically amusing, black-humor, sort of way, the exact inverse problem appears now to be going on in the neurophysiology of cognition:
Robert Hecht-Nielsen -- one of those responsible for resurrecting the field of artificial neural networks two decades after it was deep-sixed by MIT in the 1960s -- has recently come up with what he calls "Confabulation Theory" which appears to fit what is known about the neurophysiology of human cognition very well. It makes specific, falsifiable predictions about what to look for in further neurophysiology experiments and can be simulated on computers to produce human-quality natural language production. But, as he points out in the book "Confabulation Theory", a typical researcher in the neurophysiology of cognition must run a gauntlet of institutional trials which prevents him from departing from the intensely empirical bias of the field until at least age 40, by which time he has been so enmeshed in the biases of the field that to seek theoretic consilience with work outside the field -- even as closely related as artificial neural networks -- is virtually out of the question. On Thu, May 30, 2013 at 9:21 AM, Berke Durak <[email protected]> wrote: > On Thu, May 30, 2013 at 10:08 AM, Jed Rothwell <[email protected]> > wrote: > > That is completely wrong. In experimental science you never need to > explain > > how something works in order to confirm it is real. You just need to > > replicate it and show there is no error in the instruments or techniques. > > The map = Theory > The territory = Experiments > It's not on the map = No good theory > If it's not on the map, it can't exist! (Our map makers are very > good.) = It doesn't fit physics thus it's pseudo-science! (Our > scientists are very good.) > > -- > Berke Durak > >
