Reading up on Matthew Honnibal

The combination of large language models with reinforcement learning creates a path for continued improvement, making the argument that AI progress will plateau unlikely.