Project #26728 — iteration #49Minted on April 19, 2023 at 17:02
*WTBS.TXT* is the first collaboration between Thomas Noya and the Waiting To Be Signed podcast crew. Using the first year of episodes as training data, Thomas trained a GPT-2 model to generate text strings from seed words, phrases, or at random; over 3000 strings were curated from the model over several months. The strings are then randomly selected from the curated set and displayed in various fonts and patterns, evoking a retro-PC style inspired by 90s net.art.
An advantage of using so little data to train a model (yes, a year of content is nothing in AI terms) is that it allows the model to run wild and create funny and fantastical diversions. We used an AI service to transcribe the episodes, which introduced many spelling variations around fx(hash)-specific terms and artist names. All of this adds up to something that we are collectively excited to share with you.
From WTBS: A year went by fast... so fast that we fully missed doing anything to celebrate. Please think of this token as a belated celebration of the first year of Waiting To Be Signed. This project could not be possible without the hours of content we produce each week. If you have enjoyed the show, we hope you will enjoy this as well. Please collect to support the continued creation of WTBS. Collect because you are a fan (of Thomas, us or both). Collect because it is a really cool project. Huge thanks to Thomas for volunteering his time to work with us on this, and thank you to everyone who listened and supported us as we move into our second year of the show.
[s] save as PNG
WTBS ⫸ GPT-2 ⫸ p5.js ⫸ Blockchain
Copyright (©) 2023 Thomas Noya and Waiting to be Signed