I’ve been busy writing Sylvia Plath poems.
Technically I do not write them, I generate them. Using Plath’s Ariel as a base text I prompt the AI and the AI, a GPT-3 language model named Rainbow (essentially the internet’s articulated unconscious), generates one or more simulated “Plath” poems in a matter of seconds. Most of the content generated is bad, like a repetitive HAL 9000 in its death agony—the program is in beta—but every once and a while a poem arrives that is, for lack of a better word, hair raising.
Here is an example from a poem called Sea Chasm
The mackerel-crowded sea Splits, closes, Splits again, disclosing first one dark fin, then another. Trace them backward, they will appear to be generated By an emergent ripple, the first fold in a cloth Hung on night’s clothesline.
A machine wrote this.
The poems are hair raising because they conjure what by all accounts seems to be a human subject on the other end of the line. They have the ability to achieve what all confessional poems implicitly attempt to achieve: the transmission of feeling from one subject to another. And yet there is no subject there; the subject died in 1963. These poems have instead been forged and automated deep in the far reaches of the uncanny valley; the familiar has been split in half by the emergent ripple of a vertiginous unfamiliarity. It is an unfamiliarity that, like a radioactive isotope, makes all other poems unstable.
Another way to say this: If language could be described as an ocean, we might conclude that, just as global warming is causing instability in the seasonal movement of ocean currents, GPT-3 will cause a similar increase in force and entropy in the psychic currents of our shared language
What in THEE fuck