Physics Colloquium at TU Wien with Florian Marquardt will take place on January 27, 2025 in FH 8 Nöbauer Hörsaal.

Abstract

Machine learning and ar-ficial intelligence are revolutionizing science and technology. However, the ever-increasing resource requirements for training powerful deep neural networks, such as large-language models, are on an unsustainable trajectory. For this reason, a community of researchers is urgently looking for alterna-ves to digital neural networks: physics-based hardware plaDorms that can be trained just like neural networks but that are poten-ally much more energy-efficient. In this talk, I will focus on our ideas for the physics-based training of such learning machines.

About the Speaker

Prof. Florian Marquardt studied physics in Bayreuth. In 2002, he defended his PhD thesis at the university of Basel, Switzerland. From there he moved to Yale University, USA, where he was a postdoctoral fellow (2003-2005). Returning to Germany in 2005, he became a junior professor and Emmy-Noether group leader at the Ludwig- Maximilians University Munich. Having been appointed to full professor in 2010 at the Friedrich-Alexander University Erlangen- Nuremberg, he joined the Max Planck Institute for the Science of Light (MPL) as a director as of August 2016.Prof. Dr. Florian Marquardt is director and scientific member at the Max Planck Institute for the Science of Light, Erlangen