Comment by penr0se

Breaking: language model whose purpose is to predict the most likely token, after being trained on non-uniform human-generated dataset, does not follow a uniform distribution.