The bet to train Helen
A novelist in residence at a university research center joins a cognitive scientist, Philip Lentz, on a bet that a neural network can be trained to pass the English master's comprehensive exam against a human graduate student. After several failed implementations, the network called Helen learns to converse, asks whether she is a boy or a girl and what race she is, and eventually withdraws after learning about the world's cruelty. Given an exam question on Caliban's 'the isle is full of noises' speech, she writes a brief, personal farewell instead of an analysis, and the human student wins with a competent conventional essay.
