Back then, the idea that scaling deep learning would lead to significant performance gains was controversial. Several senior academic colleagues were advising me not to waste time trying to scale deep learning, but to just focus on inventing new algorithms, which was where they

→ View original post on X — @andrewyng