Unless you are just back from a vacation on Mars, you have probably heard a lot about AI, and its potential to transform the way we do things, whether that's at school, work, or even creating voices, images, and music.
There is a lot of ongoing discussion about how AI could and should be used.
YouTube has, however, recently jumped in to give us a glimpse of the future by launching an experiment targeted at its 'YouTube Shorts' community.
They are working with 9 artists: Alec Benjamin, Charlie Puth, Charli XCX, Demi Lovato, John Legend, Papoose, Sia, T-Pain, and Troye Sivan. They have licensed these artists voices to use in AI generated music tracks.
You simply type in a quick description of what the topic and style of the song you want is, and select the artist you want to perform it for you. The results are really promising, as you can see from the video below.
So you can now just ask Google to create a few songs to accompany your next routine, right?
First of all, the big issue with AI tools generally right now is that they were likely 'trained' using unlicensed data, and that opens you (and the people who created the tools) up to all sorts of copyright infringement problems in itself.
And secondly, specifically in the case of YouTube's Dream Tracks experiment, they HAVE taken the trouble to license these artists voices before training their AI tools on them. But the artists therefore have a contract with Google, and earn money for this specific use only - i.e. the use they have licensed is limited to this Dream Tracks experiment for YouTube Shorts only.
As a reminder, music is owned by its creators, and anyone else has to get their approval for their intended use, usually in the form of a commercial license. Recording artists and songwriters each have rights in how any track is used. More details can be found in our short overview of the rights required for choreographed performance sports here.
In YouTube's limited experiment, they have licensed the artists voices, but haven't yet explained how they have trained their AI on the genres of music and songwriting needed to then create the tracks.
So in future could you use AI to create bespoke songs for you?
Technically, the answer is yes - BUT there is a long way to go in terms of ensuring the full licensing provenance of the AI's used to create the track, in terms of the voices, instruments, and songwriting that was used in the training.
Beyond that, you still need to make sure that the music is licensed for the use you want to put it to. For example, the YouTube experiment is limited to a specific use - YouTube Shorts.
So it's likely that as YouTube have done, someone who can more easily negotiate with music industry rightsholders to obtain the specific licenses for choreographed performance sports - like ClicknClear! - will be better placed to provide such a service rather than it being generally available.
And would you want to?
Music is intimately linked to the human psyche, and has an incredible effect on both athlete / performer, and audience.
We wrote about these effects in our blog:
These effects are strongest when the person performing actually knows and likes the music they are performing to. And the audience (and judges...) engage more with the routine when they also know and like the music.
So until the day we all become passionate fans of the latest AI music stars, you will be better off using hit songs by human artists and songwriters!