This has been one of the more substantive and coherent conversations about ML I’ve had the chance to work through.