In the (sadly paywalled) article The Representation of Knowledge and the
Relevance of Biological Models at the Symposium on the Mechanization of Thought Processes, 1958 which appears in the journal IEEE Annals of the History of Computing, we find this note, where neuroscientist Warren McCulloch, then 60, is attacked concerning the fact that the abstract “neurons” he is presenting have little resemblance to lifeware neurons:
(N.B. the “neuron” of McCulloch is actually computing a boolean logic proposition of the true/false states at time t-1 of the neurons connected to it, it is not like the widely used “neuron” of today)
The abstract of the article states:
“Mechanization of Thought Process” was an international conference involving researchers from academia, government, industry, and the military that took place in the U.K. in 1958. It saw the first presentation of McCarthy’s Advice Taker and of Selfridge’s Pandemonium, and one of the first expositions of Rosenblatt’s Perceptron, as well as presentations on new programming languages, cybernetic experiments, and simple diagnostic systems. This article describes the conference and the occasionally boisterous debates that took place, drawing out the common challenges faced by researchers at the time, focusing on the relevance of biological models for mechanized systems of thought processing and the difficulty of embodying knowledge or context in a system to enable it to solve problems effectively. Particular attention is paid to the methodological criticisms of work in both machine translation and in what we would now consider to be artificial intelligence made by the Israeli linguist and philosopher Yehoshua Bar-Hillel.
Grace Hopper, then 51, also attended.
Further in the article, we find this passage where McCarthy’s ideas are scrutinized by Bar-Hillel
The discussion of McCarthy’s paper was opened by Yehoshua Bar-Hillel, an Israeli linguist and philosopher who was particularly significant in the growing field of machine translation and played a key role in discussions at the symposium. He had recently been commissioned by the National Science Foundation of the USA to report on the work of the various machine translation groups that existed around the world. His interventions at Teddington marked an initial outing of the criticisms he would raise in his report when it was published in 1960.
Bar-Hillel scornfully said McCarthy’s paper “belongs in the Journal of Half-Baked Ideas”(McCarthy) and attacked McCarthy’s basic linguistic assumptions (“at” was transitive, claimed McCarthy; not so, riposted Bar-Hillel), arguing that “many of the other 23 steps in Dr McCarthy’s argument are equally or more questionable.” He then spent some time attacking the lack of detail in the “common sense” (it might sometimes be cheaper to get a taxi rather than driving your car, what about the time facts and so on) and insisting that “the gap between McCarthy’s general programme (. . .) and its execution even in such a simple case as the one discussed seems to be so enormous that much more has to be done to persuade me that even the first step in bridging the gap has already been taken.”
This led to a fierce debate from the floor involving Selfridge and McCarthy, before Bar-Hillel summarized his essential point:
“I do not think there could possibly exist a programme which would, given any problem, divide all the facts in the universe into those which are and those which are not relevant to that problem.”
As Michael Genesereth has commented, “perhaps Bar-Hillel was right” (for Eliott Macklovitch, “Bar-Hillel was (almost entirely) right”). Rebecca Skinner has argued this was the first framing of what became known as the “Frame Problem”—the difficulty of finding enough axioms to enable a device to embody a sufficient and effective description of a particular environment. From this point of view, the Teddington meeting was significant both in terms of the paths that it explored and the obstacles it identified.
Here is a review by “Nature” from January 24, 1959
The Proceedings of Mechanization of Thought Processes, Volume 1 and 2, can be obtained at the Internet Archive, if you have an account:
Nautilus has an article on the collaborator of McCulloch, Walter Pitts, and his sad destiny, where we also read:
McCulloch was a confident, gray-eyed, wild-bearded, chain-smoking philosopher-poet who lived on whiskey and ice cream and never went to bed before 4 a.m.
