RELIER LA THÉORIE DE L'ÉVOLUTION CULTURELLE AUX CONCEPTS ET MODÈLES DE L'INTELLIGENCE ARTIFICIELLE
Auteur / Autrice : | Jérémy Perez |
Direction : | Clément Moulin-frier |
Type : | Projet de thèse |
Discipline(s) : | Informatique |
Date : | Inscription en doctorat le 20/09/2023 |
Etablissement(s) : | Bordeaux |
Ecole(s) doctorale(s) : | École doctorale de mathématiques et informatique |
Partenaire(s) de recherche : | Laboratoire : Institut national de recherche en informatique et en automatique - Bordeaux - Sud-Ouest |
Equipe de recherche : FLOWERS |
Mots clés
Résumé
Cultural evolution is a research field that aims at providing causal explanations for the change over time of culture, defined as socially inherited information. This field has so far remained somewhat disconnected from the artificial intelligence literature, although there seems that many insights can be gained from combining the two. First, models of cultural evolution are limited in their ability to model individual exploration. Concepts from AI such as reinforcement learning, intrinsically-motivated exploration, and goal-directed exploration could fill this gap and help us better understand how exploratory behaviors contribute to cultural transmission, innovation and open-endedness in cultural evolution. Moreover, the rise of Large Language Models (LLMs) also opens new avenues for creating more realistic agent-based models. Indeed, LLMs can be seen as tools that take cultural content as input and output new cultural content, after being trained to mimic how humans perform such transformations. Understanding exactly how cultural evolution can leverage LLMs is therefore is a very important research direction. Conversely , cultural evolutionary theory appears to be a powerful framework for anticipating the consequences of artificial intelligence on societies. Digital technologies are increasingly involved in the production, selection, and transmission of cultural content, and it is therefore crucial to analyze and predict the dynamics of machine-generated and hybrid human-machine cultural evolution. In this context, this thesis will study how 1) concepts and methods from cultural evolution research can be used to characterize cultural dynamics in populations of artificial agents; 2) how computational models inspired by intrinsically-motivated reinforcement learning algorithm can provide new perspectives on cultural transmsission, innovation and cultural open-endedness; and 3) how these new perspectives can be empirically tested with experiments and/or observational data.