AI Song Contest
Volkslied 3000 was a collaboration with Neuköllner Oper, Kultursystem and the Kommen und Gehen Festival, to foster the creation of folk music that speaks to a modern audience. We trained an AI on a vast body of German folk music texts to generate region specific lyrics based on user input. Through an open-call everybody was invited to create a composition from the AI generated texts and enter a song contest. A jury awarded the 5 best compositions, which were performed at the Neuköllner Oper and at Kommen und Gehen Festival.
Looking at the traditional German folk song from a new viewpoint requires the dissection of its definition. When is a song a 'folk song'? When it is written by the people? When it is sang by the people? When it doesn’t have an author?
VOLKSLIED 3000 conceptualises the folk song as a dynamic cultural object, an expression and embodiment of a “zeitgeist”. The project invites us to occupy a critical standpoint: Which characteristics of a tradition do we want to keep and which should we leave behind? Rather than simply preserving traditions and the outdated value systems they are attached to, how can we rejuvenate the parts of the folksong tradition that we still stand by?
At a time when algorithms have permeated social spheres, we must also address the question of whether art can also be produced by artificial intelligences beyond technical reproducibility. Does an artificial intelligence, whose linguistic ability is based on all the freely accessible texts on the Internet, capture the spirit of the times? Does it perhaps in this way already recognise a "zeitgeist" that we cannot yet even grasp?
The Volkslied3000 engine is prompt based text generator, a trained version of the artificial intelligence GPT3. The content and linguistic horizon of the generated texts is based on the general knowledge of GPT3 neural network and Wikipedia entries and a collection of folk song texts kindly provided by Volksliedarchiv.deThe AI can be "instructed" to write its own folk song texts by interacting with it through the Volkslied 3000 website.The generation is prompted by a title (free text), a year and a state to give the new texts specific characteristics.
The task of the participants was to write a song while using lyrics generated by the AI. The call was open to anybody, be it amateurs, professionals, singer-songwriters, school bands or symphony orchestras. Even though folk music may tend to have certain characteristics, it cannot clearly be differentiated in purely musical terms. Therefore we left the musical parameters completely open and left it to the creativity of the applicants to reimagine the genre.
The jury selected five winners and one special prize. Criteria for the selection were the overall artistic impression, diversity, the handling of the AI-generated text as well as the positioning on the topic of "folk song". The winners were awarded the prize money of 1,000 € each and concert possibility at the Kommen und Gehen Festival and at the Neuköllner Opera. Within the project 5 artists were commissioned to also have their take on the assignment: Max Prosa & Alin Coen, Die Grenzgänger, Tonia Reeh, João Orecchia.
The task of the participants was to write a song while using lyrics generated by the AI. The call was open to anybody, be it amateurs, professionals, singer-songwriters, school bands or symphony orchestras. Even though folk music may tend to have certain characteristics, it cannot clearly be differentiated in purely musical terms. Therefore we left the musical parameters completely open and left it to the creativity of the applicants to reimagine the genre.
All submission and the commissioned compositions are available here.
Our Role
Project Lead & Management, Concept Development, Programming
Year
2022
