YC Paper Club: Why inference is 12 times more important than training
This episode of Y Combinator’s Paper Club focuses on inference: for every 1 unit of compute spent training a model, roughly 6 times as much compute is spent serving it each year. Over a model’s 2-year lifespan, inference is 12 times as important as training. Inference optimization has created approximately 220 billion US dollars in market value over the past 10 years.
Y Combinator ·