David Ha: The Future of Collective Intelligence and Meta Evolution for Foundation Models

In this talk, I will first give a brief overview of our company, Sakana AI, an AI R&D company in Japan. Our company aims to focus our R&D directions on meta evolution, which I will describe as meta learning at the collective agents level. We will present some of our recent results from our lab, where we show how foundation models can take advantage of meta evolution principles, to achieve new capabilities relevant to the MINT workshop’s theme of improving the controllability of foundation models and developing a better understanding of their inner workings.