Everybody knows that to the unprepared reader, Joyce's novel is indecipherably chaotic.
—Jorge Luis Borges, "Fragmento sobre Joyce" (Sur 77, 61)
For a few years now, I’ve done a small essay or bit of microfiction on my blog to celebrate Bloomsday: this year, I’ve been playing around with recurrent neural networks, and decided to do one of those instead. @Usylessly is a neural network which has been trained on the text of the novel and does a passable imitation of certain features of its style, although the idea of Ulysses having a single style is a bit grotesque: one of its delights is Joyce’s chameleonic ability to shift registers, so an RNN trained on the whole book blurs these out to— something which all of these have in common? A sort of computational average of the Joycean? A style which seems less kaleidoscopic and more of its time, the further away from it we get?
I’ve often followed the various Twitter accounts which post fragments of Joyce’s works, and then unfollowed them: while it’s nice to be reminded of them, there’s something about taking little swatches from such grand texts which rubs me up the wrong way. And they make me wonder how they strike those who haven’t read the books: do they make any sense, other than as word-music? How could even the most knowledgable reader be sure that a convincing imitation of Joyce (which my bot certainly isn’t) is not in the text somewhere? @usylessly is in a way a parody of those bots: I'm not sure how patient I’ll be with it.
—Mike Lynch, Bloomsday 2016Tweets by usylessly