A step-by-step semiotic understanding of LLMs and chatbots through interdisciplinary dialogue
A step-by-step semiotic understanding of LLMs and chatbots through interdisciplinary dialogue

Summary/Abstract: Artificial Intelligence (AI) is enjoying a period of “summer”. It is an intensely discussed and researched topic which fascinates researchers across fields with questions about both its development and possible impacts. Semiotics too has immediately engaged with the topic, focusing especially on the ways in which different forms of Generative AI can be understood through existing frameworks, and how the processes of textual production it encompasses can be studied and analysed. Gen AI, however, is a complex and fast evolving technology, which is very difficult to study without a sufficient understanding of the technical side. Perceptrons, Artificial Neural Networks, Transformer Networks, Large Language Models are all terms we hear often in discussions about AI. However, they often refer to technologies we barely understand. This paper is born from a collaboration between authors with expertise in semiotics and machine learning. Our objective is to reconstruct the evolution of Large Language Models in particular, and to engage the different stages from a semiotic perspective. This in-depth engagement with the technological side is especially useful to go beyond the strong temptation to anthropomorphize the technology and instead reposition it as a tool that supports human textual creation. Based on our analysis we then propose several key concepts (an AI semiosphere, Generative AI as a mode of sign production and the concept of Intentio Machinae) as ways to map and conceptualise the role of LLMs in semiotic processes.

• Details
• Contents