Why is this recipe for capable models transformative?

In this episode, Luca and Josh chat about Big Data and The Pile, an open source resource for training models, Reinforcement Learning and Aligning Models.