IAS Amplitudes Group Meeting
How Can We Use Language Models, Here and Now?
Abstract: I will review why current large language models cannot deal with algorithmic problems without using external tools. After explaining the transformer architecture from a point of view of a physicist, I will highlight the increasing role of foundation models as an emerging paradigm for deep learning. I will suggest architectural improvements aimed to increase mathematical abilities of language models. There will be live demonstrations.
Date & Time
November 21, 2023 | 2:30pm – 4:00pm