Sam Babichenko

Reading notes

Scaling laws

Power-law learning curves and compute-optimal model size.