> a stochastic model for predicting text

It's fascinating to me that so many people seem totally unable to separate the training environment from the final product