Entre processus stochastiques et métriques d’évaluation : l’IA-créatrice à l'épreuve de l'étrangeté

Abstract

The huge amount of data produced by the Big Tech in the last twenty years has contributed to the emergence of creative artificial intelligence as an automated tool that could allegedly produce works of art ex nihilo by itself. Only distantly reminiscent of the human brain, these algorithms produce pure statistical models by sifting through millions of images, videos or songs and identifying statistical correlations that occur between their constituents. As they lack historicity by nature, these systems are only able to generate novelty by relying on random number generators. This article explores the foundations and intricacies of what appears to be a paradigm of creation based solely on stochastic processes and statistical models. We specifically examine the hyperrational perspective from which engineers envision creation, how this perspective thrives on the assumption that the post-modern subject is void of all intrapsychic conflict, how it may catch on finally in an environment where normative-referenced evaluation has become paramount.