AI Music with a Human Heart

When CTU hosted a live jazz concert at the Faculty of Information Technology, the event represented a truly unique occasion. Even though music can be considered a rather technical affair, it is not entirely common for people to hear music generated by AI algorithms. At least until now.

Cover image created by Midjourney (prompt: “jazz orchestra performing ai generated music”).

 

Like art, which we have discussed before, music is one of the areas that people do not like to be meddled with. True, music has always been linked to technology – it’s hard to imagine an orchestra without instruments, an opera without a piano, or a campfire without an acoustic guitar. But there are still humans responsible for the melody and the rhythm, even if they often create with the help of software.

The live performance by students of the Jaroslav Ježek Conservatory and College revealed to us the very promising potential of algorithms. The choice of genre was absolutely logical. Jazz attracts and demands invention and originality and thus better tolerates unusual tones composed by algorithms.

Down with monotony

The generated song itself was not perfect. In the fully electronic version, there was a certain “inhumanity”, one might say coldness, of the whole GPT-2 based composition. The algorithm also tended at times to produce sequences of notes that could not actually be played.

However, the students only had to make a few adjustments. For the live performance, it was a matter of rearranging the piece to include a sort of refrain that is pleasing to the human ear and important to our perception of music. Here, artificial intelligence has provided the menial part of music-making. Humans then gave the composition its final form.

AI jazz concert at the CTU’s Faculty of information technology, October 6, 2022. Source: CTU in Prague, Computing and Information Centre 

These are only the first baby steps into the world of algorithm-generated music, yet that makes them all the more important. For now, let’s leave out music that AI would entirely create. But imagine a musician who has a strong chorus, a song motif and good lyrics but lacks ideas on how to skillfully combine these important elements.

At that point, technology – ergo AI – could well step in.

“It’s interesting to see what is often called artificial intelligence today and where such AI can find a use,” said David Pešek, a technology scout and lecturer in entrepreneurship and innovation support, about the concert at CTU. “We are becoming increasingly convinced that artificial intelligence will not fully replace human resources but will stand in for them only in selected tasks where critical or creative thinking is not required,” Pešek added.

He summed up very well the essence of the second computer revolution, whose birth and gradual rise we are witnessing. The overarching concept of AI, artificial intelligence, does not imply a certain replacement of man by machine thinking. On the contrary. It is about supporting what humans do best – creative, imaginative, logical, and meaningful activity that will help remove monotonous, less important areas where the human touch is not needed or where human supervision is sufficient.

Algorithmic elevator music

Clever algorithms can find use where just a simple musical background is enough and songs don’t need sophistication, careful control or creative thought. For example, algorithmically generated music could serve shopping malls or waiting rooms and provide a musical underscore while holding in a phone queue.

In the future, will AI be part of creating symphonies and soundtracks for various media or music albums? It is reasonable to assume that it will. But will it create them on its own? Likely not. If only because such a creation would lack the aforementioned “humanity”. Even the best-programmed algorithm still has the limits humans give it. Nothing else is even possible with silicon technology.

AI can create simple music. It’s not that hard – services like Jukedeck or Amper Music can do such a thing already. You just have to tell the algorithm what genre, mood, tempo, instruments, length, etc., you want, and the output will be more than enough for a corporate presentation or YouTube video. It is not for professional production, though. The creations lack a “heart”.

The puzzle of authorship 

One of the key issues in AI music is legislation. Who is the author of a piece created mainly through an algorithm? Who holds the legal reins? For example, can AI be sued for copyright infringement?

The courts are already actively addressing this issue (and not just in music). At the helm of the efforts is, of course, the AI (and plaintiff) superpower – the United States of America. But the EU is also drafting relevant measures. And, as in similar cases, they will probably introduce more detailed legislation first, while overseas, they will stick to their favourite precedents.

In practice, the disputes between living and human authors show that this is certainly not a simple matter. Intellectual property decisions are not merely about technology and law, but about philosophy as well.

So who is the author, who owns the intellectual rights, and who should claim authorship? These are questions for which there is no easy answer. Ultimately, it will primarily be about consensus and perception of reality rather than tangible truth. But that’s the way it is with new technologies, and we’re just enjoying those first baby steps.

As always, there is no need to worry about the future (of music). It will be beautiful and supported by machines, not controlled by them.