Ken Aizawa brought to my attention some claims made by Alva Noe in his new book, entitled Out of Our Heads. It’s a book for the general public, so we shouldn’t expect too much scholarly rigor. Nevertheless, there is no reason why it should be as sloppy as it is (on this issue).
A section entitled “Christopher Columbus and the Brain” contains the following remarkable passage:
“For Hubel and Wiesel, … cells were understood to be specialized in order to be able to “stand for” and thus represent features. This application of information theory to the brain was not new when Hubel and Wiesel set to work… Rafael Lorente de No had represented neural relations as networks already in the thirties, and his treatment had a direct influence on the work of Warren McCulloch and Walter Pitts, and through them, John von Neumann. (Walter Freeman, the neuroscientist, likes to say that in a way Lorente de No is the godfather of the computer)” (p. 156).
In these few sentences, Noe switches from talking about the idea that the brain processes representations to the idea that information theory (presumably in Shannon’s sense, since he cites Shannon right after this passage) is being applied to the brain to the idea of neural networks to the invention of modern computers, as if all of these were just about the same thing, or at least as if you could go directly from one to the others.
As far as I’m concerned, they are all different things. For starters, Shannon’s notion of information is not representational (it’s not even semantic). To be sure, McCulloch put all these ideas together (more or less) in a theory of the mind/brain, but it’s fallacious to conflate them and anachronistic to project them back onto Lorente de No.
For the record, Lorente de No did offer evidence that the activity of “chains of neurons” play a function in cognitive processes, and McCulloch knew about that. But as Ken Aizawa reminded me, the guy who did the real work on the theory of closed loops of nervous activity is Pitts, who started working on this well before meeting McCulloch and without knowing about Lorente de No.
I know of no evidence that Lorente de No talked about neural networks, or that his putative talk of neural networks influenced McCulloch and Pitts. Instead, there is the influence of the Rashevsky biophysics group with which Pitts was working. All those guys (primarily Rashevsky, Housholder, and Landahl) were talking about and writing about “neuron networks” or “nerve-fiber networks” (since the 1930s, yes).
Bottom line, as far as I can tell: Lorente de No seems to have played a role in convincing McCulloch that closed loops of nervous activity are functionally important, but does not seem to have played a role in starting the theory of neural networks or in convincing anyone to “represent neural relations as networks”, whatever that means.
Oh, and in case you were still wondering: No, there is no meaningful sense in which Lorente de No is the godfather of the computer (with all due respect to Walter Freeman).
Of course, this comment about Lorente de No is really only made in passing and there are probably only, what, five people on earth who are professionally invested in this topic. Gualtiero and I are among those two.
Although the precise details of what Noe and Freeman are claiming is not entirely clear, the general drift seems to be that Lorente de No introduced information processing ideas to McCulloch and then into neuroscience. Maybe this is right, but I have a vague sense that things went differently.
My sense was that, in the late 1930’s, early 1940’s, McCulloch knew of Lorente de No’s work on loops of neurons. McCulloch was a neurophysiologist at Yale at the time. McCulloch moves to Chicago in fall 1941 and meets Pitts who is going great gangbusters with the mathematics of closed loops of neurons, so they write “A Logical Calculus”. At about the same time, McCulloch learns of Wiener’s work on feedback, hence comes the idea for the circular causality conferences (to become the cybernetics group). McCulloch thinks of Lorente de No as a participant, hence Lorente de No learns about information via the proto-cyberneticians. This could be wrong, but any references that might clarify matters would be appreciated.
He’s clearly a bit sloppy, but Lorento de No clearly had a major influence on thinking about neural networks. Real neural networks, not necessarily artificial neural networks.
I found this chapter to be fun, and it has a nice little intro to de No’s contributions.
Yes, I’m sort of surprised at the impressive collection of folks, such as Lorente de No, that McCulloch, Wiener, et al., were able to assemble for their cybernetics meetings. For their tenth and final meeting, they invited Albert Einstein, Bertrand Russell, and Alan Turing. All declined, but they certainly aimed high.