From the article, Gelman's money quote is this: " So I’m not knocking auto-complete; I’m just disagreeing with Hinton’s statement that “by training something to

From the article, Gelman's money quote is this: " So I’m not knocking auto-complete; I’m just disagreeing with Hinton’s statement that “by training something to be really good at predicting the next word, you’re actually forcing it to understand.” As a person who does a lot of useful associative reasoning and also a bit of logical understanding, I think they’re different, both in how they feel and also in what they do."