Also, even if we lacked the data to proceed with Chinchilla-optimal scaling that wouldn't be the same as being unable to proceed with scaling, it would just req

Also, even if we lacked the data to proceed with Chinchilla-optimal scaling that wouldn't be the same as being unable to proceed with scaling, it would just require larger models and more flops than we would prefer.