Generative artificial intelligence and new semiotic challenges

Discover the third session of the International Semiotics Seminar in Paris "Generative artificial intelligence and new semiotic challenges. Translation and creative appropriations".

This year's focus is on a specific theme among those already addressed, namely the one most closely linked to computational language models and translation problems (between human prompts and machine-revised prompts, between natural language prompts and textual descriptions of images): generative artificial intelligence.

The reflections will continue on the multiple enunciative instances involved in these intelligences, namely: the databases used to learn generative models, the algorithms used to train and infer these models, the numerical representations called “embeddings” that are created by translating visual or verbal texts into lists of numbers, the actual codes that articulate everything in a specific programming language, and finally the “prompts”, which are the queries serving as the models' entry point for the user. This approach enables a deeper focus on the many translations between different syntagmatic substances of expression and the much-discussed question of creativity.

• Philippe Gréa : University of Paris Nanterre
  From formal neurons to language models: understanding neural networks and their limits
• Adrien Deliège : University of Liège
  Dynamicity in still images according to generative AIs.