Les niveaux de compréhension des Intelligences artificielles : language, représentations et émergence du sens

Read full text Summarise with VIRAT
Authors: Daniel Defays
Cited by: 0
References: 5

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Recent developments in large language models (LLMs) have renewed philosophical debate about the nature of understanding. This paper examines the capacities of LLMs through a fourfold framework of understanding: referential (and emergent referential), inferential, pragmatic (intentional), and experiential. While LLMs clearly exhibit inferential and pragmatic capacities, and may develop internal structures supporting a weak form of referential understanding, they lack experiential grounding. The analysis highlights how language mastery alone can give rise to sophisticated forms of competence, challenging traditional assumptions that understanding requires embodiment or consciousness. These findings invite a reconsideration of how different forms of understanding relate and suggest that aspects of human understanding may also emerge from linguistic and inferential structures.

Fields of Study & Subjects

Philosophy and Theoretical Science Cybernetics and Technology in Society Embodied and Extended Cognition

Publication details

Journal / source
ORBi (University of Liège)
Publication date
22 March 2026
Language
EN