A small group of us are working on an NLP project and we are looking for some feedback! We have fine-tuned GPT-2 on a dataset of poetry and we are using a generative-adversarial type environment where a classifier is trained to identify human from neural poetry.
We would like a human baseline to assess whether humans are more adept at spotting the artificial poetry than our discriminator and it would be greatly appreciated if you could fill out our short quiz!
Our approach to building the poetry generator was heavily guided by Gwern Branwen’s work which is great (https://www.gwern.net/GPT-2).
TLDR: If you have a spare couple minutes then have a go attempting to classify poetry as either human or AI written!
AI poetry quiz: https://agolajko.github.io/