AI can drive our cars, can predict our chances of dying of a certain disease but these are purely logical uses of AI. The algorithms are used to make logical decisions based on previously acquired data. With art such as music, decisions are not purely logical as it involves emotion. Can AI therefore really make music that will be enjoyable? Will AI end the human art of music, or will it inspire new musicians and even start a new wave of creativity?
Don’t think that AI is new to the music industry. There are multiple examples of AI being used support in the music production process. There are programs which help during the mixing phase of music to make it sound good on multiple outputs, such as car speakers, different types of headphones and different streaming services like Soundcloud or YouTube. This saves a lot of time for the producer, since a track doesn’t have to iteratively be rendered, listened to on the desirable output/platform, and be adjusted.
However, there are also examples of AI applications that are able to produce music on their own. One of them is Amper, an application which produces music by pre-defining certain parameters such as genre, tempo, mood and key [1]. With this program, people are able to produce stock music fast and easy. To see how good the resulting music was, the company ran a Turing test which tests a machine’s ability to exhibit intelligent behaviour to, or indistinguishable from, that of a human [2]. The results showed that consumers could not tell the difference between music composed by Amper and music composed by humans [1].
The main criticisms of AI music is that it will end the creativity in the music industry as all music will start to sound more or less the same. The main downside of musical AI’s, is that they are limited to previous data and thus older music. Moreover, it is difficult to program human creativity in an AI, since we don’t exactly know how creativity works in the human brain. It is like trying to explain the difference between a dog and cat to a computer. We as humans know exactly how to distinguish them, but even explaining it to other people let alone programming this in a computer program, is still difficult. On the other hand these criticisms might also be the fear of ‘the new’, as was the case when the use of synthesizers came up in the 70s, or when Hip-Hop music started to become popular in the 80s/90s. Music did not die, on the contrary, it allowed for new creativity to flow which lead to new interesting and exciting music.
It is difficult to predict the future of AI in music. We are still at the very beginning of this development. We have examples of creation of songs by dead artists such as Frank Sinatra [3]. This is possible due to the immense music catalogue that we have. As you can hear, it is still something that has to be developed more, but we can certainly expect more of these applications. Imagine Tupac and The Notorious B.I.G. on the same song, 25 years after their death. However, it will be difficult to include the emotional aspect of music creation, which will be the main limitation. Maybe this will be possible in the future, but the moment AI’s can show human emotion will be special not only for the music art, but much for other applications such as human like robots. Until then, we will have to enjoy Frank Sinatra covering ‘Toxic’ by Britney Spears.
Listen for yourself, can AI win the next Grammy for ‘Album of the Year’?
- https://www.youtube.com/watch?v=4MKAf6YX_7M Blue Jeans and Bloody Tears, a Eurovision song created by Artificial Intelligence
- https://www.youtube.com/watch?v=gA03iyI3yEA On the Edge – AI Generated Rock Music Composed by AIVA
- https://www.youtube.com/watch?v=iJgNpm8cTE8 Never Gonna Give You Up, but an AI attempts to continuously generate more of the song Don’t worry I’m not rick rolling you
- https://www.youtube.com/watch?v=03xMIcYiB80 Romanticism in D minor – AI Composed Music by AIVA
- https://www.youtube.com/watch?v=a8ewkIe9fq4 Travis Scott AI Robot (Jack Park Canny Dope Man)
- https://www.youtube.com/watch?v=hH-7emvgo0o Jukebox AI regenerates “city of stars” using Frank Sinatra’s voices and music style.
- https://www.youtube.com/watch?v=v4kevbEuSSQ Artificial Intelligence makes Frank Sinatra sing “Toxic” by Britney Spears
References
[1] https://time.com/5774723/ai-music/
Thanks for your interest blogpost about AI in the music industry! The added YouTube links show how the technology can be used to combine two artists from different generations. Quite amusing of course, but not the quality music people want to listen to. Spotify figured out how to use AI for improving the listener experience. Artists know this and adjust their music to Spotify’s algorithm to be ranked on a desired position in the Spotify chart. Accordingly, artists try to combine their creativity with AI, but is this for the better? To be honest, I think that the top-50 charts show that we as consumers and artists as producers lost sight of what real creativity is. obviously, you have to commercialize your music to make a living out of it, but there will never be a next rolling stone like band if you keep supporting this way of working.
The use of AI is inevitable, but do you think that it will ever be possible to combine this with creativity? In my opinion, I believe that it is important to shift the power of platforms to the artist and reward them for their creativity in a more personal way. Do you think that this will be possible? Or is this quality lacking development inevitable?
Hi Mels, Your example of Spotify shows the main flaws of the new technological developments in the music industry. You can write a separate blog just about Spotify. Have you noticed how songs have become much shorter recently? Most songs don’t last 3 minutes long, a lot of songs even end around the 2,5 minute mark. The reason is the business model behind Spotify and its recommendation algorithm. This probably coincides with the decreasing attention span of the new generation(s), but that’s another discussion. I believe the music industry has to undergo some drastic changes in how it operates, and how artists are compensated. The music labels and corporations have all the power, and they decide what most people will listen to and what they will like. Luckily, with the emergence of YouTube, streaming services and other social media, new artists are able to be more independent. I think we are at the beginning of a major shift of how this industry operates, which will shift the power from the corporations back to the artists.
I agree with you that artists need to be rewarded for their creativity. I believe this will not happen until the music industry and its corporations and labels will be regulated, which will give the power back to the artists. As for using AI, I believe artists will find ways to creatively use it. It will be difficult to create emotional music purely by using AI. We as people can immediately hear when something sounds emotionless and bland, so I don’t think we will listen to AI generated music. Music is more than just the music. It is also the people behind it and their emotions, thoughts and motives.
Here are some interesting reads on the streaming industry:
https://qz.com/1519823/is-spotify-making-songs-shorter/
https://www.theverge.com/2019/5/28/18642978/music-streaming-spotify-song-length-distribution-production-switched-on-pop-vergecast-interview
Nice subject to write about! 🙂 I really see AI changing things in the music industry, but personally I think that AI will probably be more of a “helping hand” for musicians, rather than taking over the music business. Especially what you write last makes me put this statement like this. Emotions are very important in the creation of music and AI will never have emotions themselves. They might one day, be able to show emotions as they learned to mimic us well enough, but I kind of feel like I wouldn’t dig music the way I do know, knowing that there is a computer behind it, instead of a human being like we are.
Btw, the videos were fun to watch, but music created by AI is definitely not there yet. 😉
This was a very interesting read! I may be a bit more optimistic on the capabilities of AI, but I do believe that, one day in the far future (but perhaps less far then we assume), AI will be capable of creating music that feels as if its composer unloaded his/her emotions into it.
I don’t think that we have fully grasped yet what human creativity entails, but it seems to exist somewhere in the realm of associative thought processes, which I assume are somehow regulated (or at least influenced) by emotions.
Creativity, assuming that it does come down to associative thought processes, should (eventually) become computable with neuromorphic architectures (e.g. Intel’s Pohoiki Springs).
To me, the interesting question (implied in Joost’s comment; thank you!) that underlies all of this is the following: in the context of AI-generated music, at what point could emotions become computable parameters, if ever?