It’s an incompletion that numerous experts in the world of classical music have tried their hands on. Schubert’s ‘Unfinished Symphony’. Almost two hundred years later, Huawei, in collaboration with composer Lucas Cantor, has done the impossible: completed the symphony using Artificial Intelligence in a smartphone.
Franz Schubert never completed his Symphony No. 8, only writing two movements and sketches of a third. Now the Huawei Mate 20 Pro smartphone has done what scores of composers have attempted for almost two centuries. The final two movements of Schubert’s eighth, completed by Artificial Intelligence, were performed on 4th February in London’s prestigious Cadogan Hall by the 67-piece English Session Orchestra. Far from being robotic and unfeeling, the symphony surged into a soaring, epic finale. This memorable and emotional experience triumphantly beckoned in the future and a new era of music. It was an incredibly hopeful moment for music and technology – which left us with more questions than answers.
This achievement was the result of a collaborative effort of artificial intelligence and human genius – in this case Lucas Cantor’s.While Huawei’s smartphone churned out hypothetical melodies, it was Cantor’s role to bring the emotion and context in a way that only a human could.We spoke to the Emmy-winning composer to illuminate the music process and tell us how with the help of a smartphone he completed what Schubert couldn’t.
SARAH OSEI Can you describe your role in completing the symphony?
LUCAS CANTOR You’re wondering whether I’m the composer or not? No, it was a collaboration between myself and the Huawei Mate 20 Pro – the AI. What we did was we spent a few months teaching the AI to sound like Schubert. The AI, like a human composer, needs to learn, and we had to curate the information that it would learn from. Our goal was to teach it what the DNA of Schubert was so it could generate melodies that Schubert mighthave written but didn’twrite. I listened to those melodies, I chose the ones that I thought were the most appropriate and arranged them. In some cases I added to them and in some cases I minimised them. It was truly collaboration the way I would collaborate with another composer. Using the AI was like having a collaborator that never gets tired, doesn’t have a bad attitude, isn’t afraid of bouncing ideas off me. So it was really amazing to have this sort of tireless sidekick by my side as I wrote this piece. To put it into a literary analogy: the AI provided me with some clever terms of phrase that were consistent with the style of the original author – in this case Schubert – and I wove those into a story.
SO How was this different from working on your own?
LC It’s certainly the perception that the classical composer slaves away in a room by himself and shows up at an orchestra rehearsal with music, but I think it has always been a collaborative effort. This is a collaboration between Huawei, the producers, myself – the composer -, the orchestra and the conductor. Music has always been this way: what brings it alive is many people coming together to make it happen.
SO There are a lot of people who see these kinds of developments in a pessimistic way, as technology encroaching on creativity. When you were first approached with this project, were you sceptical at all about the place of technology in making art?
LC There have always been those people. When the phonograph was invented they thought it would get rid of every musician’s job, and that didn’t happen. When Pro Tools was invented they thought it was going to revolutionise the industry and put a lot of people out of work – and that didn’t happen. I think this is going to be another tool that I use that is going to help me be creative and help me generate music. Maybe it’ll help me generate it faster, maybe it’ll help me generate it in ways that are different than what I would have thought of on my own, but I don’t think that the robot armies are going to be taking over orchestras in the future – ever. I know for a fact that is not Huawei’s goal. What we believe is that it’s collaboration between the Artificial Intelligence and the human intelligence that makes beautiful art. It’s about augmenting and helping me, the human, express what I want to express in a different way.
SO You said a few times that the work was quicker, I can imagine that’s an incredible asset as a composer.
LC Yes! If Schubert could write faster he might have finished the symphony himself. If he had access to the technology that I had access to he might have been able to be more prolific than he already was. While this project has been in development for several months and we’ve been talking about it since the autumn, the writing process started for me about 30 days ago and what you’re going to hear tonight, did not exist on January 1st, even conceptually, for me. And it’s about 18 minutes of original music – depending on how fast you play it (laughs). To write 18 minutes of concert music in a month is an accomplishment.
SO Were you surprised by the results?
LC I was really blown away by the results initially. First of all, I collaborated with non-human intelligence on writing a symphony, that in itself is very profound. What did surprise me was that I initially thought that it was going to generate melodies that were more or less interpretations of other Schubert melodies of the period, and what it didgenerate were melodies that were a little bit more modern, but not outside the realm of 1822. The reason for that – going back and studying Schubert’s other symphonies – is this symphony was really different. It’s technically not like classic symphonies of the time. It is quite uncommon, and I think had Schubert lived longer and had a longer career, he would have come back to this piece 20 or 30 years later and finished it in a more modern style. It was profound realisation that he was going for something that he was not able to attain in 1822, and that’s why I think he put it aside. So what I had to do when I got those melodies was honour that style. Each of the movements that I wrote have melodies that were generated by the AI which I thought were beautiful and really evocative. Those melodies were written in a way that I would not have thought of on my own.
SO How much of the emotion in the completed piece is your work?
LC This is the question that I keep getting “did you bring the emotion to the piece or did the computer bring the emotion to the piece?” and my answer is that the audience brings the emotion. I write with whatever feelings I’m writing with but the way that you experience it is up to you. It has nothing to do with me, it has nothing to do with the smartphone.
SO Did you use Schubert’s sketches for the third movement?
LC We did not, no.
“I think Schubert didn’t finish this work because he ran out of ideas and had he had an AI he might have just finished it himself.”
SO So this is quite a new approach, this is a revolutionary moment in music actually, what were the challenges or problems that you encountered?
LC It is a revolution but it feels to me that it’s not that different to projects that I’ve done in the past where I was getting input from a source that was unique, but at the end of the day I was getting input in a form that I recognised – in the form of melody. and I was able to use that input to create a symphony. The analogy that I’ve been using is this: Neill Armstrong didn’t walk very far to get from the ship to the moon, but a lot of work went into getting him all the way up there. So from my perspective, I just took a step, but the team at Huawei, the team developing the AI, did a lot of that work to get me there.
SO As this technology develops further, do you expect AI could create original music without human assistance?
LC I love this project because these questions get so philosophical and these are the questions that I’m asking myself. What I can say is that this technology helped me write a symphony relatively quickly and to imagine what that might be doing in a few years or even in a few months is amazing. To answer the question that I think you’re trying to ask: do I think I’m going to be replaced by a computer? I don’t think so. Part of what this project was about was collaboration between human intelligence and Artificial Intelligence and how that could generate something that was different than what AI would generate on its own and also different than what I would compose. So if Huawei had asked me to finish Schubert’s symphony and then hung up the phone, I could have done that, I could have done something but it would have been completely different than what you’re gonna hear tonight. The fact that we’re collaborating with this AI has steered the project in a direction that is unique and quite beautiful.
SO How would you like to incorporate this kind of technology in your career from now on?
LC I’ve always been interested in collaborating in unique ways. The next symphony I’m going to write is about the Arctic ice shelf. There are some scientists that have been able to record the sound over the course of a year or two, and then compress those vibrations into smaller sound waves from which they can extrapolate data. I’m working with one of these scientists to take some of those years of data and use AI to see how this would sound like over a thousand years and what kinds of interesting sounds and music we can get out of that. So I would love to use AI in the future, I think I will. I hope that it will be a tool in the composer’s toolbox. As a writer sometimes you’ve got a blank page infant of you, you brainstorm, right? Or you just sort of start and usually it will jumpstart the engine, and I think AI can do that for you. Like I said, it’s like having a collaborator that’s not afraid to bounce ideas off you and is not afraid just to start, which is something which I think humans struggle with.
SO Was it stressful to work with a collaborator that’s always working?
LC (laughs) No, because I could tell it when to stop. It was really quite interesting, I started feeling a kinship with this machine. There were moments that I was frustrated and didn’t know where to take it, but it did help to have this collaborator. I found it really useful to get an endless stream of ideas without ego. If I got to a point where I didn’t know what to do I had so much information that I could draw from and I found that really helpful.I think that it’s possible that Schubert didn’t finish this work in particular because he just ran out of ideas and didn’t know what to do and had he had an AI he might have just moved on and finished it himself.
Interview SARAH OSEI