ChatGPT: A mathematical perspective

Prof. Dr. Yann Disser, TU Darmstadt, Optimierung

With the release of ChatGPT large language models have quickly become an important tool in many intellectual occupations, including academic teaching and research. But we are just beginning to tap into the breakthrough potential of this technology and we can expect its importance to further grow dramatically in the near future. The promises and dangers of accessible artificial intelligence have sparked discussions in the mathematical community about the future of teaching and even research. This talk intends to provide a rigorous footing for these discussions by explaining the technology behind large language models and showcasing its capabilities. More precisely, we will establish mathematical basics of deep learning and the general architecture of language models, briefly discuss the current state-of-the-art, and show applications of chatbots like ChatGPT in mathematical teaching and research.