Sensi senza significato. La competenza semantica al tempo dei Large Language Model

The question of «meaning» proves to be a divisive and challenging subject for AI. Scholars and professionals in the field of Large Language Models have been debating this issue for years without finding a consensus.The disagreement lies not only in the interpretation of results but also in the theoretical assumptions that should lend scientific validity to discussions about models. This paper aims to provide a unified theoretical framework for understanding «meaning» in AI systems like the renowned ChatGPT. The proposed framework is based on Saussurean structural linguistics, which proves effective in describing model functioning and resolving many of the issues raised in specialized literature.