Speaker
Description
This contribution reviews the historical roots of contemporary Artificial Intelligence, highlighting its connection to both twentieth-century Statistical physics and Cybernetics.
In the 1920s, within the field of statistical physics, the well‑known Ising–Lenz model was introduced. Over the course of the twentieth-century, it gained increasing relevance and also found applications in “non‑traditional” areas of physics. In the 1980s, with the introduction of Hopfield networks and Boltzmann machines, the statistical connectionist approach progressively gained ground, replacing the logicist‑symbolic paradigm typical of classical Artificial Intelligence, ultimately receiving recognition with the 2024 Nobel Prize.
Alongside this line of research, in the first half of the twentieth-century, Norbert Wiener (1894–1964) and others proposed, within the study of intelligent behavior, the new science of cybernetics: “the science of control and communication in the animal and the machine”. In Italy, cybernetics found particularly fertile ground within the community of physicists as for example, the Neapolitan cybernetics school founded by Eduardo Caianiello (1921–1993), which in the 1960s hosted N. Wiener as a visiting scientist.
Many contemporary AI techniques, besides being connectionist, make use of concepts originally developed within cybernetics. Consider, for example, the links between Transformers (used in Large Language Models) and Information Theory, or between Convolutional Neural Networks and Signal Analysis.
In this light, contemporary AI, emerging around the beginning of the twenty‑first century, is better understood as rooted in the statistical physics and cybernetics research originating in the first half of the twentieth century, rather than as evolution of symbolic‑logic AI from the second half.