Generative AI and Semantic Grounding
Projektleiter:
Finanzierung:
Haushalt;
The project aims to explore how core features of modern Generative AI systems pave the way to the semantic grounding of such systems. It centers around the following working hypotheses:
- Generative AI, including LLMs, is based on generative models.
- Generative models are probabilistic structural representations that act as world models.
- Generative models are key for self-supervised learning.
- Semantic grounding is a gradual affair that unfolds along the three dimensions of functional, social and causal grounding.
- Causal grounding scales with the development of world models.
Kontakt
Prof. Dr. Holger Lyre
Otto-von-Guericke-Universität Magdeburg
Fakultät für Humanwissenschaften
Institut III: Philologien, Philosophie, Sportwissenschaft
Zschokkestr. 32
39104
Magdeburg
Tel.:+49 391 6756574
weitere Projekte
Die Daten werden geladen ...