Making a Mind vs. Modeling the Brain: AI Back at a Branchpoint, by Hubert L. Dreyfus and Stuart E. Dreyfus | in mind <> computer

"informatica" appears to be an interesting magazine, around since 1977, and published by the jožef stefan institute in ljubljana. at least in the issue from 1995 stories about fights over funding in science and whose ideas win (apparently) was published. it's a story of the early days of ai, their concepts and drawbacks (ie "parallel distributed processing", physical symbol system, ...), embedded in the old question of how to mirror mind.

... The rallying cry of the first group was that both minds and digital computers were physical symbol systems. By 1955 Allen Newell and Herbert Simon working at the RAND Corporation had concluded that strings of bits manipulated by a digital computer could stand for anything - numbers of course but also features of the real world.

the attack, 1965. minsky and papert made their position clear with the statement:

Perceptrons have been widely publicized as "pattern recognition" or "learning" machines and as such have been discussed in a large number of books journal articles and voluminous "reports". Most of this writing... is without scientific value. | minsky and papert, perceptrons p.4

and some rhetoric tricks, by just talking about the one-layer perceptron, even, so dreyfus, they were quite right – partly.

Yet in the conclusion to Perceptrons when Minsky and Papert ask themselves the question: "Have you considered perceptrons with many layers?" they give the impression while rhetorically leaving the question open of having settled it.

Well we have considered Gamba machines which could be described as "two layers of perceptron." We have not found (by thinking or by studying the literature) any other really interesting class of multilayered machine at least none whose principles seem to have a significant relation to those of the perceptron... We consider it to be an important research problem to elucidate (or reject) our intuitive judgment that the extension is sterile.

Their attack of gestalt thinking in A.I. succeeded beyond their wildest dreams. Only an unappreciated few among them S. Grossberg J.A. Anderson and T. Kohonen took up the "important research problem". Indeed almost everyone in AI assumed that neural nets had been laid to rest forever. Rumelhart and McClelland note: Minsky and Papert's analysis of the limitations of the one-layer perceptron coupled with some of the early successes of the symbolic processing approach in artificial intelligence was enough to suggest to a large number of workers in the field that there was no future in perceptron-like computational devices for artificial intelligence and cognitive psychology.

... Holism could not compete with such intense philosophical convictions. Rosenblatt was discredited along with the hundreds of less responsible network research groups that his work had encouraged. His research money dried up he had troubled getting his work published he became depressed and one day his boat was found empty at sea. Rumor had it that he had committed suicide.

Whatever the truth of that rumor one thing is certain: by 1970 as far as AI was concerned neural nets were dead. Newell in his history of AI says the issue of symbols versus numbers "is certainly not alive now and has not been for a long time". Rosenblatt is not even mentioned in John Haugeland's or in Margaret Boden's histories of the AI field [note: deals with the 1977 book, margaret boden has another vol. out by now, yet i don't know, if rosenblatt got included].