Atlas record: https://www.aitimeline.dev/en/events/evt-0041/
Research · 1969
Marvin Minsky and Seymour Papert published a mathematical analysis of what a perceptron can and cannot do.
The book established concrete limits of single-layer models. It is often linked to the decline of neural-network research, although it does not claim multilayer learning is impossible.
How far the book affected research funding remains a matter of historical debate.
MIT Press · Published 1969
Advances in Computers, vol. 37, Academic Press (copy hosted on gwern.net) · Published 1993
Minsky and Papert in 1969 were answering the programme set out in this book.
The book formally analyses which problems a single-layer perceptron cannot solve.
Both texts assess the expectations of the time critically, each in its own area.
A network with feedback connections sidesteps the limit proved for the single-layer scheme.
Aleksander wrote in 1983 that neural modelling in image recognition suffered a quasi-death in 1969 with Minsky and Papert's condemnation; he presents WISARD as a practical answer.
It refutes the broad reading of the perceptron critique: with a hidden layer the limitation disappears.