speculative: what Pitts was working on, what others did, and what might have been · an essay from the house

Speculative. The first two parts are documented. The third is a guess. Nothing in it happened.

I. what is known

he 1943 paper (McCulloch & Pitts, 1943) specified every connection in its nets. Four years later McCulloch and Pitts asked how a brain could recognise a shape whatever its size or position (Pitts & McCulloch, 1947), and the answer they gave was already statistical: average over many transformations. After that, Pitts's interests moved from nets whose every link is chosen to large three-dimensional nets whose links are random, and to what such nets would do on average. Gefter describes his doctoral work in these terms [paraphrase, not a quotation]. Wiener, who had made the mathematics of random processes his own, was the obvious supervisor.

The thesis was never submitted. Pitts burned it, with as much of his other work as he could find (Smalheiser, 2000)(Lettvin, J. Y., 1999). No draft is known to survive. See the break.

II. what others did instead

The questions did not wait for him. Cragg and Temperley (1954) compared assemblies of neurons to the atoms of a magnet. Beurle (1956) treated the cortex as a mass of cells passing waves of excitation. Caianiello (1961) wrote equations for the dynamics of large nets. Wilson and Cowan (1972) described interacting populations of excitatory and inhibitory cells. Hopfield (1982) brought the statistical physics of magnets to memory in nets, and helped restart a field that had stalled.

Authored, not historical. III. what might have been. This part is a guess, and is labelled so.

Perhaps a thesis on the statistical behaviour of nets, written at MIT in the early 1950s with Wiener's mathematics behind it, would have arrived a generation early at what Beurle, Cowan and Hopfield found later. Perhaps not.

Two things argue against the romance. The frog work of 1959 (Lettvin et al., 1959), which Pitts helped do, had already shown that the eye does not hand the brain simple inputs to be combined by logic, and it seems to have shaken him more than any letter. And the field's later answers came from many people at once, which suggests the ideas were in the air rather than in one man.

What is lost is not a result. It is a person who might have spent forty more years thinking, and the conversations he would have had.

Authored, not historical. In the tour this essay has a room. The bound thesis on its desk has a blank spine; visitors who have read this page find a title pencilled on it, "On the statistical behaviour of nets". That title is invented. No thesis of that name exists. The docent's card in the room is ours too: "We kept a desk for it. We keep the lamp on. It is not a vigil. It is only that nobody has told us to turn it off."

in the tour: the room called "The thesis that wasn't" has a placard called "What might have been" ···

notes

1. Parts I and II cite the works linked. Part III, the invented title and the docent's card are ours.

Sources

1. McCulloch & Pitts (1943). A logical calculus of the ideas immanent in nervous activity. The Bulletin of Mathematical Biophysics. doi:10.1007/BF02478259
2. Pitts & McCulloch (1947). How we know universals: the perception of auditory and visual forms. The Bulletin of Mathematical Biophysics. doi:10.1007/bf02478291
3. Gefter, A. (2015). "The man who tried to redeem the world with logic." Nautilus, February 2015. nautil.us/the-man-who-tried-to-redeem-the-world-with-logic-235253/
4. Smalheiser (2000). Walter Pitts. Perspectives in Biology and Medicine. doi:10.1353/pbm.2000.0009
5. Lettvin, J. Y. (1999). "Pitts, Walter." In R. A. Wilson & F. C. Keil (eds.), The MIT Encyclopedia of the Cognitive Sciences. Cambridge, MA: MIT Press. ai.ato.ms/MITECS/Articles/lettvin1.html
6. Cragg & Temperley (1954). The organisation of neurones: A co-operative analogy. Electroencephalography and Clinical Neurophysiology. doi:10.1016/0013-4694(54)90008-5
7. Beurle (1956). Properties of a mass of cells capable of regenerating pulses. Philosophical Transactions of the Royal Society of London. Series B, Biological Sciences. doi:10.1098/rstb.1956.0012
8. Caianiello (1961). Outline of a theory of thought-processes and thinking machines. Journal of Theoretical Biology. doi:10.1016/0022-5193(61)90046-7
9. Wilson & Cowan (1972). Excitatory and Inhibitory Interactions in Localized Populations of Model Neurons. Biophysical Journal. doi:10.1016/s0006-3495(72)86068-5
10. Hopfield (1982). Neural networks and physical systems with emergent collective computational abilities. Proceedings of the National Academy of Sciences. doi:10.1073/pnas.79.8.2554
11. Lettvin et al. (1959). What the Frog's Eye Tells the Frog's Brain. Proceedings of the IRE. doi:10.1109/JRPROC.1959.287207

see also

	the break
	Walter Pitts
	timeline: neuroscience