A brief note on the methodology: due to the fact that Talk to Transformer will only admit a small block of text as a prompt, and will only write a few sentences afterwards, I must assemble these stories by repeatedly cutting-off the top few sentences, and then running that block of text through as a new prompt, to generate a few more sentences; I then piece them all back together. No deletions, no modifications of any kind — this is pure robot imagination.

Yet, I do have a few criteria for pruning: I run the prompt through again if the outcome is repetitive or incoherent (this happens often, where it will take one phrase from the prompt and repeat it endlessly; weird bug); otherwise, I paste it to a word document, and run again. After one or two dozen runs, I pick the one that is most in context and interesting. Oh, I also skipped the weird sex scenes, references to common fictional characters, (the neural network loves Harry Potter and Star Trek!) as well as intermittent racism. The AI learned everything from us. :/