Can a linguistic model understand? Logical and philosophical questions about the semantics of LLMs

Generative artificial intelligence (AI) using large linguistic models (LLM) is capable of generating content on a par with that of humans in terms of originality, completeness, and logical coherence in the articulation of concepts. Integration with automated content generation systems is progressively changing knowledge practices and transforming the way we interact with knowledge and the underlying logic.

This question brings us back to the semantics of natural languages, their architecture, and how this new technology can be appropriately integrated. The conference aims to explore the similarities and differences between natural and automated languages ??and to understand the potential and dangers associated with their use.

9:00 Introduction and start of work

9.05 Vittoria Dentella (UNIPV): LLM and grammar

10.05 Chiara Manganini (UNIMI): A pragmatic theory of LLM errors

11.05-11.30 Break

11.30 Guido Seddone (UNISS): About the metalinguistic skills of LLMs.

12.30 Giuseppe Primiero (UNIMI): Functional responsibility in AI

1.30pm Lunch break

15 Federica Malfatti (University of Innsbruck): Learning from the machine? Reflection on the role of AI as an epistemic mediator

16 Silvano Zipoli Caiani (UNIFI): The reasons for AI: reality or fiction

17 Pause

5.30pm Claudio Paolucci (UNIBO): Machines equipped with language: the myth of meaning.

6.30pm Federica Porcheddu (UNISS): Grammar of constraint and archive of potential.

Discussants: Fabio Bacchini, Marinella Cadoni, Stefano Caputo, Massimo Dell'Utri, Silvano Tagliagambe

Organizers: Ludovica Lorusso, Guido Seddone

Info: lorusso@uniss. it

European Union funding - Next Generation EU relating to Ministerial Decree 737/2021 2012/22 resources