Generative AI and foundation models have had spectacular success in text and image processing (e.g., GPT4, DALL-E3, SORA). We believe that modern AI can play a similar role in the sciences and address some of the long-standing problems in Physics, Chemistry and Biology. Compared to text and images, there is comparatively little high-quality data available in the sciences and therefore model design, inductive biases and built-in physics play a greater role. In this talk I will introduce to some of the long-standing problems in molecular simulation and discuss recent approaches using generative AI and foundation models in this space.
