The Netherlands - COMPUTD / Shuman&Angel-Eye

The core team members of COMPUTD went to university together in Maastricht. Their mission is to make AI and data sciences accessible and understandable. The team is supplemented by producers René and Angela Shuman, who participate to investigate what a collaboration between AI and people can lead to. The goal of the team: to have AI write a catchy song with minimal human intervention.

What makes a song catchy? During a brainstorming session with the entire team, they set several preconditions for the machine-learning algorithms. It had to be a cheerful song, so all songs with a negative sentiment were filtered out of the database. The team wanted to prevent their song from becoming a copy or collection of samples from existing Eurovision songs. For this, the outcome of a possible melody was always compared with the original dataset. Did it sound too much like an already existing song? Then the algorithm had to make further calculations.

The subject of the song is simple, but effective: AI writes a song. Initially, the team thought it was wise to filter the dataset for English-language songs and generate lyrics by syllables, while "rewarding" their algorithm if it had generated the words 'AI' and 'writing a song', but syllable generation turned out to be not so easy. In the end, individual pieces of the lyrics were created by AI, and the team manually shifted the lyrics, chords and melodies into a coherent whole. Production wise, the sound image was carefully selected, with the team trying to summarize what AI intended and delivered.

Team COMPUTD / Shuman&Angel-Eye is very satisfied with what they have achieved in their free time with five people. Their process confirms that AI cannot (yet) create the feeling and depth that characterizes a person's creativity. This requires an important guiding creative human touch. René and Angela Shuman: 'AI could be a great addition to people in the future. Not the other way around.'

René and Angela Shuman own ALWAYS-ONLINE BV, an IT company that provides cross-media productions for international brands such as TESLA, AUDI and PROFILE. As performing Artists, they are known for the many TV productions they made for various broadcasters.

COMPUTD's three university friends are big music fans, but they also each have their own expertise. Dirk Bongers has experience in bioinformatics, medical data and predictive analytics models.  Pieter Schaap is active in both IT and the world of games and has a lot of experience with stock market trading bots, algorithms used in games, and deep learning.  Marcell Ignéczi specializes in finding patterns in data and text, which came in handy when analyzing Eurovision songs and lyrics.

Everything they make serves one goal: to provide something useful to the largest amount of people possible. Whether it’s a software package or a Eurovision song.