Recently, the US Copyright Office (USCO) ruled that an ‘AI’ (Artificial Intelligence) cannot claim copyrights itself. Stephen Thaler claimed this on behalf of his ‘Creativity Machine’. This AI had created a series of artworks based on machine learning. Mind you, it was the AI who claimed the copyright in this case, not the creator of the AI itself. The USCO ruled that the Creativity Machine could not claim copyrights on the images produced because the AI lacks one important element, namely being ‘human’. In other words, only real people can own copyrights.
What applies to pictures also applies to music. Aimée herself cannot claim any copyrights on the music she produces. There’s good news for YouCompose users: Aimée doesn’t claim it either.
But what about the creator of Aimée, Wout. He has copyright on Aimée, but can he have copyright on the music composed by Aimée? After all, there are a lot of creative hours in the programming of Aimée and as far as we know Wout is human.
Legally, this has not been established unambiguously because there is no case law on this (yet). But there is good news for users of YouCompose and Aimée: Wout, as the author and owner of Aimée, will not claim any copyrights on Aimée’s music.
Who is left then? Well, you! As a user of YouCompose, and therefore of Aimée, you can
claim copyrights to the music composed by Aimée. After all, it happened on your initiative and you chose the parameters. There is always a ‘human’ component here. You used Aimée, to a greater or lesser extent, as a tool in writing your music in the same way a photographer uses his camera or a keyboardist uses his synthesizer.
But is the latter correct? Is Aimée comparable to a camera or a synthesizer? When granting a copyright claim, the USCO places great emphasis on the fact that a work of art must be the product of human creativity. When a person uses a camera or a synthesizer this is not disputed, but does that also apply to Aimée?
This is not yet clear in a legal sense. The coming years will show what level of human input is sufficient for the USCO, or another agency that deals with it, to assign the copyright to the produced artworks to the AI user.
For now, in any case, as far as Wout is concerned, you own the copyrights to the music that was composed on your initiative by Aimée on your computer, iPhone or iPad.
But… consider this:
During one of the many test runs of YouCompose, Aimée came up with the following melody:
You may recognise in this melody a resemblance to Beethoven’s 5th symphony. Now Beethoven can no longer claim copyrights on his music because he passed away more than 70 years ago, but suppose he didn’t and Beethoven still had copyrights on his music. Beethoven could challenge your copyright claim for this composition, although Aimée composed the above piece for you completely autonomously and without knowledge of Beethoven’s music.
There is one more aspect that is important. For this we return to the above-mentioned
‘Creativity Machine’ by Stephen Thaler. This makes use of ‘machine learning’.
This means that the AI, before it is able to create something itself, is ‘trained’ on the basis of existing works of art by one or more authors, the so-called ‘data’. Many contemporary creative AI systems work in this way.
Recently, there has been increasing resistance to this in the creative community.
Artists worry that their works, which are sometimes difficult to create, will be used as data in ‘machine learning’ AI systems that then create new works that are suspiciously similar to their creations.
Two aspects related to intellectual property and copyrights play a role here:
- Can a certain artwork be used to train an AI?
- Does an AI generated artwork infringe on the copyrights of a human artist, in a similar way as in the above Beethoven 5th example.
The two aspects have not been tested in a legal sense.
Regarding the first point: many AI makers who build systems based on existing artworks invoke the principle of ‘fair use’ of copyrighted material. Human artists would therefore do well to indicate for each work whether or not it may be used as ‘data’ in machine learning systems. When an artist or composer explicitly forbids this, I think there is no longer any question of ‘fair use’, but as I said, this is nowhere laid down in a legal sense.
As for the second point: Any work of art, whether man-made or machine-made, can infringe on someone else’s rights. So anyone can dispute the copyrights of an AI-produced work for whatever reason. A judge will ultimately have to determine whether the claim is successful.
What about Aimée and the ‘data’ she uses? Well, Aimée does use data, but not existing music. Aimée’s data consists of music-theoretical knowledge about melody formation, harmonization, instrumentation etc. brought together in an expert system. Wout added his own knowledge and consulted well-known sources that are included in a bibliography that you can find in the settings app for YouCompose and although the books in the list are of course copyrighted, the knowledge described in them is not. Combining this knowledge with genetic algorithms creates a personal Aimée style, so that you can recognise Aimée’s compositions as you often can with other composers. Aimée can do her work without using copyrighted sources and is thus a truly autonomous composer.
It may be different if you, as a user of YouCompose, use Aimée as an assistant and have a piece composed based on pre-existing and possibly copyrighted melodies. In that case, the work created by Aimée may also infringe someone else’s rights. Keep that in mind.
The basic principle is that you, as a user of YouCompose, own the copyrights of the music produced by Aimée on your computer, iPhone or iPad. Neither Aimée nor Wout can or will contest your rights. Aimée herself does not use copyrighted material of others.
It is possible that the result of Aimée’s efforts infringes other people’s rights. This could be coincidental or the result of the fact that you put Aimée to work on the basis of melodies that are protected by copyright.