My friend who would like to remain anonymous made a zine called "What Will Happen to Me." She used the GPT-2 neural network to generate 100 answers to the prompt "What will Happen to Me?" and selected her favorites for the zine. I scanned some of the pages to share here.
From the zine description:
"GPT-2 is neural network software that completes the beginning of a story you tell it. GPT-2's database contains the text of millions of web pages, which it uses to produce the stories, one word at a time, in the same way smart phones use autocomplete to help people write text messages. The stories it generates contain the interests and cultural biases of the internet-using population.
Because the software includes an element of randomness, it will generate a different story every time you enter the same prompt. I told GPT-2 to complete 100 stories beginning with "What will happen to me?" I chose my favorites, lightly editing them by removing redundancies and other artifacts."
What I noticed immediately was the dark, morbid nature of many of the AI generated responses. If the entire internet was a single brain, it would be a very tormented one. Although some of the responses don't make complete sense or flow the way that a normal sentence would, I think this adds to the eeriness of the zine.
