Former xAI Founder and University of Toronto Associate Professor Jimmy Ba: More Reporting Layers Mean More Lossy Compression