NYT goes deep on Google AI.

Link. “you can represent a language pretty well in a mere thousand or so dimensions”. I remember Hopfield’s neural net excitement in 1979.