The discussion of Large Language Models reminds me of Joseph Conrad's insulting remarks about language students and Russians in "Part First" of "Under Western Eyes".
Did Conrad like anyone?
"To a teacher of languages there comes a time when the world is but a place of many words and man appears a mere talking animal not much more wonderful than a parrot"
...
"The illogicality of their attitude, the arbitrariness of their conclusions, the frequency of the exceptional, should present no difficulty to a student of many grammars; but there must be something else in the way, some special human trait—one of those subtle differences that are beyond the ken of mere professors. What must remain striking to a teacher of languages is the Russians’ extraordinary love of words. They gather them up; they cherish them, but they don’t hoard them in their breasts; on the contrary, they are always ready to pour them out by the hour or by the night with an enthusiasm, a sweeping abundance, with such an aptness of application sometimes that, as in the case of very accomplished parrots, one can’t defend oneself from the suspicion that they really understand what they say. There is a generosity in their ardour of speech which removes it as far as possible from common loquacity; and it is ever too disconnected to be classed as eloquence.... But I must apologize for this digression."
gutenberg.org/cache/epub/248…
There are many ways to draw the distinction -- none of them fitting neatly into a Twitter post -- and there is a lot we don't understand well.
But “stochastic parrot” is the wrong way to do it. It suggests that what fundamentally distinguishes LLMs from human brains is that the former are probabilistic prediction engines and the latter are not. Yet the most influential models of the human brain describe it as doing something strikingly similar. slatestarcodex.com/2017/09/0…
Sep 26, 2026 · 10:21 AM UTC
1
3
2
221

