To that end, Bertram compares the output of two models: first, the standard GPT3 Davinci model, and second, Bertram’s custom model named Warpland, a version of GPT that they have trained on the corpus of Gwendolyn Brooks. In the second, middle section, which puts GPT’s output side-by-side with Warpland’s, GPT’s biases are evident: in response to “tell me a Black story,” GPT reproduces clichés: stereotypical stories in which Black people suffer until they pull themselves up by their bootstraps and triumph.