NeuroAI: Language Representations in Humans and Machines

Language comprehension in humans involves intricate neural networks that process meaning through both written text and spoken words. Prof. Dr. Fatma Deniz investigates these processes using advanced brain imaging techniques like functional MRI to map the regions of the brain activated during natural language tasks. Her research reveals that semantic processing is represented in similar brain regions regardless of whether language is heard or read and offers a deeper understanding of how context influences language representations. This talk bridges neuroscience and artificial intelligence, examining how findings from brain research can contribute to explaining and optimizing large language models. By comparing representations in the human brain with AI-generated language representations, the talk highlights key differences and similarities in how meaning is constructed across modalities and languages. Beyond explainability in AI, these insights have practical applications for treating language disorders and developing AI systems with human-like language capabilities.