Pop music made by actual robots is here… and it sounds considerably better than you might think.
Hello World, released earlier this month via Flow Records, is being touted as ‘the first multi-artist commercial album created using Artificial Intelligence’.
The LP has been recorded by French collective SKYGGE, in collaboration with the likes of Canadian chart-topper Kiesza and Belgian pop star Stromae… and, of course, those all important computers.
SKYGGE is led by composer, author and producer Benoit Carré, alongside a gentlemen who is becoming increasingly well-known (and slightly fretted about?) in music business circles: François Pachet.
Pachet (pictured) is the world’s foremost scientist in the field of AI-assisted music creation. Aka: Music composed by machine minds.
Last summer (in news that got a few tongues wagging amid the service’s ‘fake artist’ controversy) Pachet joined Spotify as Director of the platform’s‘Creator Technology Research Lab’.
His recruitment by Daniel Ek’s company followed 20 years of service at Sony, where Pachet – a semi-professional musician in his own right – pioneered projects which resulted in the first known pop songs composed with AI.
In late 2016, we got to hear those songs: Daddy’s Car, in the style of the Beatles, and the less conventional The Ballad Of Mr Shadow, in the style of American songwriters such as Irving Berlin and Duke Ellington.
Sony made no bones about calling these tracks, created by Pachet’s Flow Machines division, the “first ever entire songs composed by Artificial Intelligence”.
(In reality, they actually needed the expert assistance of Benoît Carré to arrange the music and write the lyrics.)
Thing have progressed considerably since these germinal prototypes – so much so that Hello World contains its very own mini-hit.
Hello Shadow, feat Kiesza, has racked up over 1m plays on Spotify to date, having been added to no less than 15 of the streaming service’s localized New Music Friday playlists.
The potential of machine-created music has some in the entertainment industry excited, yet others are flat-out skeptical – even spooked – by the idea of CPUs superseding human-driven composition.
Exhibit A: Soon after MBW broke the news of Pachet’s arrival at Spotify last year, one senior major label A&R confidentially mailed us – in a missive rife with anguish – to remark: “Can you imagine a Glastonbury singalong to a computer-generated chorus? Words fail me.”
His other comments were rather less printable.
For Pachet, a semi-professional musician, such concerns have little bearing on what he sees as his central mission: manipulating the power of AI into tools which push artists and songwriters into exciting and uncharted creative territory.
And, very possibly, changing the shape of the worldwide music business while he’s at it…
how do you feel the songs on ‘Hello World’ stand up as a body of work, if you were a reviewer?
I’m happy you asked that question. We want people to listen to this [album] as a piece of music – not as a technological artefact of some sort.
I like these songs very much. I am very often a bit jaded by music – I’ve been listening to music for a long time, I’m a huge Beatles fan – so it is very rare for me to get excited by music these days. But I am very, very excited by all 15 songs on this album.
When I listen to music I always listen for new stuff – what’s new, what’s strange, what’s compelling, what’s outside of conventional stuff.
How do you respond to people who say these songs will endemically have a lack of soul, because they are at least partly a product of machine thinking and machine intelligence?
I say to them, please first listen to the songs before you form an opinion about them. I don’t see any lack of expression or emotion in this music.
I can see a lack of expression or emotion in many [other] technical demos that people do all the time, that’s for sure, but here we have made what I consider to be real music.
I don’t think [a perceived lack of soul in the music] is a weak point here.
Should we as a society and as a music business be fearful of what you are developing and what this ultimately means for the future of music creation?
I think we should be [feeling] exactly the opposite: we should be really excited.
The way I see it, the only reason [I’m doing this] is because I believe it’s going to open up new ways of making music. It’s going to change music, bringing fresh material, new ideas, new melodies, new chord harmonies, new chord progressions, and also new sounds and new ways to combine sounds.
It’s going to allow many people to create new stuff.
“AI is going to change music, bringing fresh material, new ideas, new melodies, new chord harmonies, new chord progressions, and also new sounds and new ways to combine sounds.”
That [apprehension] is something that happens sometimes in the history of technology and music, when a new instrument is born.
When the piano was invented and the saxophone, and then in the 80’s you had the digital synthesisers which was completely crazy and a new revolution.
To me [AI] is a bit comparable to that, in that in a sense it’s just a new generation of tools.
The subtle difference being that the piano doesn’t write and play itself, and the digital synthesiser doesn’t create its own composition. Do you foresee an era where a computer with no human input can create a song that is the most popular song in the world?
I don’t think that’s a very realistic scenario, no.
Already, for a very long time, we have had algorithms which are able to generate complete [musical] pieces. That’s nothing new – that was achieved in 1958. But these algorithms are not yet going to do anything all by themselves; they need a musical message, an artist who has something to say and a curator of the technology.
“The way I see it is it’s more like assisting the Lennon that needs a McCartney.”
It’s not just a one button thing [to create a song]. That’s not the case today and I don’t think it will be the case soon – maybe it will never be the case.
This is really about empowering people and giving them ideas. The way I see it is it’s more like assisting the Lennon that needs a McCartney, or vice versa.
We are building companions and collaborators who are smart enough to give good ideas [to humans]. But they are not sufficient to create it all on their own.
Who owns the music produced by artificial intelligence? And is the answer to that question a threat to artists, songwriters, publishers or record labels who benefit from copyright?
It’s a very complex question. I’m not a legal guy, really – but what I can say is what we are doing with the album, Hello World.
[We posit that] all the songs have been created by artists using this technology, and so our composers are credited the standard way.
The tools are never credited [with copyrights]. Maybe you would have to credit the trumpet of Miles Davis in that case!
“I think [AI] is an opportunity for everyone in the music industry, and the first people to benefit will probably be the artists.”
As for the labels… well labels are needed to manage the rights and to exploit the assets.
I created an independent record label, Flow Records, to distribute and promote this album – and I hope many more to come.
I think [AI] is an opportunity for everyone in the music industry, and the first people to benefit will probably be the artists.
I don’t see how labels could not benefit from [AI] because it will create more music and hopefully better music.
How will the tools you are building specifically benefit songwriters, especially in context to what you are working on Spotify?
I joined Spotify very recently, so it’s very early to talk about what we are doing.
For the moment we don’t have much to say, but the long-term plan is to develop AI technologies that really are at the service of artists; that’s the primary goal.
I think this is just the beginning of a new era, basically.
Is the goal to allow or assist more people to enjoy the magic of becoming composers of musical works, and do you think that’s a good thing?
I think it’s always a good thing when tools democratise [music] in some sense.
Again if you take the comparison with synthesisers, they opened up for a lot of people the possibility to compose and create their own music; those people could not [make music] before and I feel that’s quite a good thing. I guess the same will happen with AI.
Personally, I am more interested in working with artists, in particular, singular artists; people with something special to say.
I want to see how these technologies can challenge them and help them move a little bit away from their comfort zone. That’s the current vision.
One scenario I can see is that when a pop single becoming successful, AI is used to replicate it, and the music industry becomes more homogenized.
I can understand that fear, or maybe that fantasy. But all of these arguments forget one thing: eventually, who decides that the music is good? It’s the listeners, right?
So if you create bad music or if you create music that’s a copy of some existing stuff, after a while people will get bored – they will not like it.
I think we have to trust in the power of listeners.
Listeners want things that are new, interesting, compelling, exciting and stimulating.
I don’t believe at all that this or any technology will contribute to uniformization of music.
You’ve mentioned there has to be human input today for a meaningful track to exist, but in principle do you think artificial intelligence could autonomously create moments of music that touch humankind emotionally?
That’s a very complex question.
Suppose I play you a song that you have never heard before, and it’s great. How can you say it was not generated by a machine? What do you even mean by ‘generated by a machine’?
Currently a machine is not able and will never be able to generate anything completely by itself. Machines require data – music for instance – in order to do new things, exactly like humans listen to music a lot before they [compose] new things.
“There are things that machines are just very bad at understanding.”
It’s not so much the process of creation from scratch that is at stake here – this is more about, How do you help people metabolise, understand or do something new with whatever they have in their heads?
I don’t believe at all that machines will ever be able to make anything from scratch [alone]. I don’t see that soon, or even in a very long term for many reasons, including technical ones.
There are things that machines are just very bad at understanding. Take the social aspect of music – when you are a musician you know about society, you know what people listen to currently today.
The machine has no idea about all that, its not really able to cope with this information. So I don’t see the question as being grounded in reality.
You were hired in a role in Spotify which no-one was doing previously. this is a brave new world for employer and employee. Why do you think Spotify hired you… and why do you perhaps think a music label or publisher didn’t think to hire you?
They hired me because I’m very smart [laughs]!
My official title is not secret at all – I’m director of the research lab. So I’m not creating a product; I’m conducting research on AI, technology, and how it can be useful for creation.
[AI-assisted music] is not a new subject, again it was invented in the ’50s or ’60s, but because of the recent progress in AI there are lots of new exciting things to study.
One of my fantasies is to participate in the creation of a new musical style. I believe that is possible; that’s why I’m here, basically.
“One of my fantasies is to participate in the creation of a new musical style. I believe that is possible; that’s why I’m here, basically.”
I am not saying I will succeed, but it is a target of mine.
I’m not a business guy; I’m a researcher, and I’m a musician who would like to contribute to the evolution of music. I would like to be as excited by music as I was when I was young and a Beatles fan.
I think technology can help re-create an era of high creativity where many new things are done all the time – both new ideas, and new musical excitement.Music Business Worldwide