>However, the complexity of our probabalistic word machine is far greater, in terms of both richness of inputs, motivation, and dimensionality.
If thought (as expressed in language) is just probabilistic pattern matching, then how did we develop our own training data from scratch?