Models currently have well enough data, orders of magnitude more than humans get. When’s the last time you read the entirety of Wikipedia cover to cover? The mo

Models currently have well enough data, orders of magnitude more than humans get. When’s the last time you read the entirety of Wikipedia cover to cover? The models are just currently too dumb to make use of it properly. It should be possible to train a more intelligent (if less knowledgeable) neural network with a fraction of current training dataset. The reason that they use so much data is because it’s easy gains.