A.I. Music - State of the Nation
- David
- Jul 25
- 4 min read
Updated: Aug 6
At the end of 2023 we posted a blog Using AI Music In Your Sports Routine?
Things have moved on significantly in the last 18 months in this fast-moving field.
Fortunately, we haven't got to the stage of AI replacing athletes just yet... but how are we doing on AI replacing musicians?
Legislation
'Big Tech' are continuing to push for unrestricted (i.e. free) access to content upon which to train their ever improving generative AI models which can create high fidelity images (like the one above), videos, sounds, and other types of content.
They are having some wins in the courts - a Federal judge in Northern California recently ruled that Anthropic's use of over 7 million copyrighted books to train its AI models without their authors consent was 'fair use' (a US legal term meaning that brief excerpts of copyright material may, under certain circumstances, be quoted verbatim for purposes such as criticism, news reporting, teaching, and research, without the need for permission from or payment to the copyright holder).
This was the first of many ongoing US copyright lawsuits brought to give an answer about fair use when it comes to generative AI. It doesn't, however, mean that other judges will automatically follow suit.
The UK government's 'Data (Use and Access) Bill' - which was generally supposed to be about improving online access to data in order to grow the economy and reduce bureocracy - met with fierce opposition from the creative industries including music (Elton John and Dua Lipa!), film, and TV who felt it could be used to legitimise 'theft' of their copyrighted material. Despite this, it eventually passed and is on its way to becoming law, although a separate A.I. focussed bill is planned which will likely provoke even stronger arguments.
The Music Industry
In the music industry (and creative industries generally), there is a strong negative sentiment against generative AI being used to replace human creatives.
Music industry lawyers are busy suing AI companies, whose products they accuse of being able to regurgitate whole chunks of copyrighted recordings - at the current count Universal and Sony are claiming $350M US dollars damages from two AI companies, Suno and Udio, whose services can allegedly do this for around 2,500 copyrighted songs.
At ClicknClear, we have seen a case where an athlete ripped music from a social media site which was allegedly created by AI but actually contained whole choruses of very famous songs that were indistinguishable from the originals. This also emphasises the point that using music 'ripped' like this doesn't in any way provide the set of permissions that athletes need from the rightsholders of such music to use it to accompany their routines.
On the flip side of the coin, the music industry is also exploring ways that AI can be used in a curated way to help the creative process. For example Warner Music Brazil recently used AI in collaboration with Brazilian musician Cleber Augusto, whose throat cancer took his voice 21 years ago. The voice of a 'donor' singer who has similar characteristics to Augusto was used as the foundation, and AI did the rest to produce new material in Augusto's voice and name, with his enthusiastic consent.
"The voice is the soul of music, and AI allowed me to maintain the essence of my interpretation, even with technology. The end result is amazing,” he said.
"The emotion and message of the songs remain intact, and that is what really matters. I thank Warner Music and everyone involved who made this project possible."
Things are moving fast - all the examples referenced in the blog are from the last few weeks.
It is still early days, and governments, the law, and the music industry are scrambling to catch up with the issues raised by generative AI.
Just because we can, does it mean that we should?
The music industry will likely settle into a system of licensed control of AI trained on their copyrighted material, and leverage it for the benefit of music creators whilst continuing to pursue outright regurgitation of works.
Ultimately, music is hardwired into human DNA, and using music that represents "who you are" positively affects you as an athlete, judges, and audiences alike.
Consider the implications of the image at the top of this blog, and in the video above of Boston Dynamics carefully programmed dancing robots - once you move past the initial 'hah' of seeing robots perform your sport, what would be left of the human expression and experience involved in competing - would judges score the coding involved? And who would want to watch that?
Similarly with music - consider Cleber Augusto's words, focussing on the soul, emotion, and message of his music. Would you, your audience, and judges be inspired by AI generated music?
The Velvet Sundown, a 'music project' presenting as a human band that got over 800,000 listens on Spotify was revealed to be AI-created, and after the initial hype seems to have sunk without trace. Who would become passionate about their music?
In a move that perhaps illuminates the motives for The Velvet Sundown and others that may follow, the people behind it said it was AI-generated but "guided by human creative direction". In other words they are claiming the copyright on the AI's output, and want to be paid for it - it's a quick and easy way of 'making a buck' for them.
Our thought therefore currently remains as in our previous blog: copyright law won't go away and fundamentally whoever - or whatever - creates and owns the music you want to use, you will still need permission (a license) before using it.