Pierre Marquis - A Guided Tour of Artificial Intelligence Research: Vol. 3 Interfaces and Applications of AI

This Springer imprint is published by the registered company Springer Nature Switzerland AG

The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland

Artificial Intelligence (AI) is still a relatively young research area (just over half a century), but its media coverage has been massive since its inception, sometimes accompanied by polemics. Today, there is hardly any, and the AI research has been broadly diversified, as shown by the three volumes of which we have the last. This variety necessitates a historical perspective: This is what I present in this preface.

As early as 1946, the French publisher Hermann published Norbert Wieners Cybernetics or Control and Communication in the Animal and the Machine, which was a great editorial success. And in 1954, an International Congress of Cybernetics (the first one) was held in Namur and I had the chance to participate in it. Artificial Intelligence has only begun to be mentioned as a research area in its own rightor even an autonomous disciplinein the mid-fifties (1955), in the famous Dartmouth Report, written by John McCarthy, Marvin Minsky, Nathan Rochester and Claude Shannon, the founding fathers of AI. It will be followed by numerous others, as well as by congresses and colloquia on Artificial Intelligence (the baptism of which is a little later). The two domains were obviously neighbors and were often confused, and indeed the interdisciplinary boundaries with other domains such as automation, computer science and cognitive research still remain rather vague today. A vast array of AI topics and subtopics has finally developed. A synthesis like the one presented in this book is therefore welcome.

I had the privilege of being involved in the early developments of AI. First at the Atomic Energy Commission (CEA) where I headed the Analog Calculation Laboratory and then at Euratom where I had set up a Research Group on Automatic Scientific Information. The subjects we discussed at this time were, mainly:

For obvious economic and political reasons, the subject of machine translation has benefited from the considerable investment but was abruptly reduced when it became apparent that our knowledge of linguistics (especially about semantics) was still insufficient to lead to acceptable results. In any case, the algorithms that we developed (the objective of which being the development of anti-combinatorial procedures) required computational resources that only a few teams could then have at their disposal. Progress was therefore rather slow, and the impatience of the decision makers was rising. It was reinforced by media excesses in which too many promises were expressed that were not kept. It is interesting to read again the popularization book, written in 1952 and published in 1953 by the NRF in the collection Lavenir de la science, directed by Jean Rostand. This book was entitled La Pense Artificielle and titled Introduction la Cyberntique. The author, Pierre de Latil, was a scientific journalist whom Norbert Wieners book had strongly impressed. Latil, in his book, emphasizes a mechanism whose research during the Second World War, especially those relating to antiaircraft defense, had shown the importance: the concept offeedback. For this concept, which is at the center of Wieners analyses, Latil suggested a happy neologism in French: rtroaction (p. 52). The concept of feedback wasimplicitly or explicitlythe basis of many automata projects that were then proposed with Ashby, Gray Walter, McCulloch, etc. The enthusiastic Latil calls these projects revolutionary, despite the modesty of their results and impact. In fact, this impact is essentially on the media and arouses old myths such as the Golem that Wiener will later evoke in God and Golem, Inc., MIT Press (1964).

A significant example of the problems that arose in the early 1960s, as well as the directions taken by research, was presented in the talk I gave at the second AFCALTI conference, the French Association of Automatic Computing in 1961. This talk was entitled Encountering numerical and non-numerical problems in the elaboration of a program dedicated to the resolution of the game of Go-bang (p. 221 of the report). It elaborated over the theme and problems of the formal representation of situations, and their evaluation, that of the paths of the trees in which they occur and of the strategies they allow to construct. We were moving from the digital era to the symbolic era.

The non-digital programming research line, which began this way, represented an important turning point in the development of electronic computing machines, previously devoted to the scientific computation which dominated space and nuclear engineering. The non-digital theme was widely discussed at the IFIP Congress on Artificial Intelligence held in Munich by Marvin Minsky in 1962. I presented a paper entitled Research on Artificial Intelligence in EURATOM, but these issues had also been discussed at the two symposia held shortly before by IBM in Blaricum (the Netherlands) with the participation of John McCarthy and many logicians. Some of the papers presented there appeared in the book I edited in 1963 with David Hirschberg at North Holland, which was published in the series Studies in Logic and the Foundations of Mathematics, Computer Programming and Formal Systems. In particular, one can find in this book an important contribution by John McCarthy: A Basis for a Mathematical Theory of Computation (p. 33) which is at the origin of the LISP programming language, as well as the famous article by Chomsky and Schtzenberger: The Algebraic Theory of Context-Free Languages (p. 118). The book I published in July 1968 at the Presses Universitaires de France (PUF), in the collection La Science Vivante, directed by Henri Laugier, under the title Intelligence Artificielle gave an update on this research results. The first of its kind, he was featured as such in the