Podcast with Eduardo Miranda, Composer and Professor of Music

My guest today is Eduardo Miranda, Composer and Professor of Music at the University of Plymouth. We talk about what quantum music is, how quantum computers help in composing and performing music, how to turn my violin into a Stradivarius, and much more.

Listen to additional podcasts here

THE FULL TRANSCRIPT IS BELOW

Yuval: Hello, Eduardo and thank you for joining me today.

Eduardo: Hello. It's a pleasure to be here. Thank you for the opportunity.

Yuval: So who are you and what do you do?

Eduardo: Okay. Well, sometimes I introduce myself as a composer with a background in computing and some other times I introduce myself as a computer scientist with a musical background. I've studied both topics at college and at postgraduate levels. And I have been working professionally on both camps. Most of the time at the intersection of art and science. And currently, I am a Professor in Computer Music at the University of Plymouth in the UK.

Yuval: My first encounter with Computer Music was when I was a kid, I went to the Computer Science Museum or Computer Museum in Boston. And there was an exhibition by Ray Kurzweil, it was an automatic accompanist. So, you could play the violin and then the piano part would automatically follow you. But I don't think that's exactly what you are working on. By the way, if such a product existed, I would buy it on a heartbeat. So tell me about your work. What are you composing? What is quantum music?

Eduardo: Okay. I've been at this intersection of music and computing since I started this studying at college. Probably because I'm not a very talented performer, I started to use computers to make sounds for me. And so I've been programming machines to synthesize sounds and also to help me to come up with ideas for compositions for the last 30 years or so. And this relationship of composers within computers is a long standing one. Since the 1950s, composers having been exploring computers to compose. And it's a natural progression, I think, that composers will inevitably start exploring the possibilities of quantum computing in musical composition and other musical tasks as well.

So, for the last three years or so, I've been immersing myself, first to learn about quantum computing, it's not an easy thing to do as a self-taught person. It's a completely different mindset from classical programming. But I got there eventually. And, at the moment, I'm exploring how quantum computing ideas may harness my creativity and hopefully I should be able to formalize these thoughts in ways in which I can pass on my discoveries to other colleagues and musicians that might be interested in exploring this as well. So, yeah,. So that's basically what I'm up to these days.

Yuval: So, what is the quantum computer composing? Is it the melody? Is it the accompaniment? Is it the exact sound or harmonics? What part is done using a computer and what part is done with a human?

Eduardo: Okay. This is a multi-faceted question. There are many ways to answer this. My answer is what I am doing. I've been always interested in programming machines to produce materials for me to compose with. So these materials maybe be either synthesized sounds, which I then take to the studio and work with the sounds, shaping the sounds to achieve particular musical ideas and so on. This is what the so-called electro-acoustic music composers do; it is purely electronically composed sounds in music.

And I also program machines to generate musical materials, which can be saved as MIDI data, which is a form of encoding music, which I can then upload into musical editors to see these scores, and then I can work with these scores to craft my music. So because computers intrinsically are generators and processors of data, the art of composing with computers is to translate ideas into data in a format that computers can process for you, and then translate the results from this processing back into some form of musical representation that you can work with.

So, computers basically produce patterns and data for me, which I then develop ways to translate into musical forms, which I can then take to the studio, take to my workstation in my office and so on, and craft these into musical compositions. I am not so keen of the idea of using computers to replace human creativity. AI can do wonderful things like compose pastiches of known composers and so on. But I'm more interested in using computers to harness creativity, to come up with ideas and materials that would probably not have been possible to do manually. And I think this is the wonderful thing about technology and the era we are on, especially now with such different emerging [quantum computing] technology that is coming.

Yuval: Where does the quantum fit in? Most of what you said until now, I would think could be done on a classical computer as well. Let's ignore the number of qubits required and so on, but in terms of algorithmically, if I wanted to build a state machine that says Do Re Mi Fa So La Si, what's the next note, is it a Do or is it something else, I could probably build it classically. Where does quantum come in?

Eduardo: Correct. You are absolutely spot on. That is what my research is about. What is it in quantum algorithms, the quantum way of thinking, that could bring a different way of doing things? At the moment, what I have been doing is to see whether I can extend what has been done classically in the quantum realm. I want to learn, I'm learning, I'm in a learning process at the moment where I am experimenting with very traditional ways of generating music using machines, but translating these processes into quantum processes.

Now I've been learning a number of things for example, that there are bottlenecks in classical machine learning algorithms for music, which makes learning very slow. And this, if you want to learn musical vocabularies, musical styles, or ways of generating music by looking at how people do it, it takes a long time for classical machines to do that. So you would not be able to kind of work with computers on the spot on a real time performance situation. You would have to train machines to do that beforehand.

But I'm beginning to realize that perhaps if we have hybrid machines where some form of quantum processing can speed up this learning task, we may be able to interact with the machine in much more sophisticated ways. Where instead of learning, for example, as we do at the moment, we have one machine that is looking at one musician performing and interacting with that single musician alone on the stage. We may have a situation where we may have an orchestra that is being watched by a machine. And this machine is then producing responses at that level, at the level of the orchestra composition.

So, things can become more complex and you may be able to process much larger amount of data in a much shorter period of time. And I think this is exciting from a musical performance and composition point of view. So that's only one of the aspects, where people can ask, well, “what is the advantage of quantum over classic?” But to be completely honest with you, I’m not just thinking terms of advantages, I'm also thinking in terms of a different approach, different ways of composing. And this is what I am very excited about in my research.

Yuval: I think it's fascinating that a lot of the quantum terms have musical analogies. For instance, when you think about superposition, that's harmonics in music. When you think about interference, that could be the reason that if I play C and C sharp at the same time, it sounds bad.

Eduardo: Correct.

Yuval: So, I think the foundation is there.

Eduardo: It is, if you think of the wave function. Now for a musician that is very natural to think in terms, now of a sound is a wave function, where you have many components with different amplitudes that is shaping the timbre. So, a timbre is, or can be thought of as a quantum state.

Yuval: What is the desired output? Will the quantum computer generate sheet music and musical score that someone an orchestra or a group can play? Or is it really just sound that could be either standalone or together with a human performer?

Eduardo: I think both are valid pathways to explore these. Because I am a composer that I have been working with contemporary classical music and I've been composing for orchestra and so on. So I'm interested in this orchestra side of things, especially, I'm interested in using machines to help me to orchestrate. To help me finding the right combination of instruments in an orchestra that will give me particular timbres of particular effects and so on.

This goes back to what we just talked briefly about thinking about timbre as a wave function. So, if I can give the machine a sound and then say, "machine, please give me the orchestration that will give me him a sound that sounds like this," that would be one wonderful and these are the sorts of things that I'm looking at. But having said this, the idea that you can use computers as musical instruments, as synthesizers is also very attractive. And I think there is a lot of potential to have some sort of quantum musical instruments that you can play in real time. I think what kind of sounds it can produce, or will produce, that is up to us to dream of and build these things.

Yuval: So,  if you think about the arrangement for orchestra, one of the famous pieces is Pictures in Exhibition. Mussorgsky has the piano score, and then Ravel did the orchestration. You think one day we'll see a computer generated orchestration of the Pictures of Exhibition.

Eduardo: Yeah. I have no doubt about this, no. This can be achieved today with classical computing. But I think the problem here, in the bottom of the problem, is that there are so many combinations that you would take ages for a human being to go through all these combinations, and find the ones that are suitable, the ones that would work. Even in terms of the, you know,  to figure out that a flute cannot play a note that's too low or a violin cannot play certain intervals that are impossible for a player to make. So these are the kinds of knowledge that traditional classical AI doesn't have because it's too much, too much information to put in the system. Whereas if we have an unlimited possibility of processing power, we can put all this kind of information in a system and the system can do these sorts of things for you in a much more optimized way.

Yuval: So that ties the quantum machine learning, being able to say, "Well, this is how Berlioz did it, or this is how Brahms would do this orchestration-"

Eduardo: Correct.

Yuval: ... Sorry, I'm going back to the classical period but, and then say, "I want an orchestration in the style of this person."

Eduardo: Correct. I think that is the next frontier, I think, for computer aided composition. For now, what we've got is pitch generators, right? So a computer generates notes that people often play on a piano, or synthesizes for a piano, and then the composer goes there and orchestrates this and makes the arrangements and so on. But I think now we can begin to think about going to the next level. Now, also generates timbre, also generate combinations and perhaps even be able to have sorts of catalogs of emotional things that you may want to achieve with the music.

As a musician yourself, you may know that, Berlioz, wrote a treatise of orchestration, where he says, no, he tells composers, "If you want a melody to sound happy, you combine these and these instruments. If you want it to sound aggressive, you combine such and such instruments." So he wrote a recipe for composers to achieve specific moods with the resources of the time, of course. But these kind of things now can begin to get automated and I think the quantum computers will enable us to get more sophistication on these automations. That's my wish anyway.

Yuval: Forgive my ignorance, but has a quantum computer piece been performed or recorded by a professional group? Are there recordings available where the music was composed by a quantum computer?

Eduardo: Well, incidentally, yes. I've been doing this on more experimental level. I composed a piece which I entitled Zeno. Which is for a bass clarinet and electronic sounds. So what I do was, I wrote the bass clarinet part, the performer plays the notes and so on. And then, my laptop listens to the notes and represent these notes in a way, which I can encode them as a states, in an array of qubits. And then, I send this to a quantum machine over the cloud. I get the qubits measured with some quantum algorithms that I designed and I get the results back. And these results are then converted into sounds that are played alongside the bass clarinet. Of course, I have to account for the delays in transmission and conversions and so on, but this is all part of the of the composition. So I've got a recording of it which will be available for the listeners of these podcasts.

And at the moment, I'm working on a major piece [Multiverse Symphony] where I'm going to put together all the arsenal of algorithms and experiments that I've been developing, which is a piece for an orchestra of 10 musicians. And there will be some of the instruments will be connected to a computer on the stage connected to a quantum machine,  which will also be converting the materials they play into a format that is suitable for feeding quantum algorithms that will be processing the sounds and the music and generate responses in real time at the concert. So this is a major piece for orchestra, which I believe it might be, I don't know, it might be the first of doing this. I'm always cautious to say, “this is the first”, but it may be a unique, let's say, piece of music where I'm going to put into practice, the number of things that I've been developing.

Yuval: It sounds just like a professional musician or professional athlete. It takes a lot of work and effort to make it sound effortless.

Eduardo: I've been working on this, I would say that this composition has been on the makings for the last three years since 2018, 19 when I began to explore quantum computing. I probably unconsciously started thinking about how I can put quantum computing into practice, into a composition, a real composition. And I think the time now has come where this composition will have to be made. And I started actually writing it two months ago. So I've got two pages already, but it's a lot of effort. But that's the joy of it, I think that's what a composer does.

Yuval: Absolutely, and so as we get close to the end of our conversation, I just wanted to ask you about timbre or pitch and so on. Do you think that a quantum computer can help, for instance, make my violin sound like a Stradivarius, could change the composition of the tone so it sounds like a higher quality instrument?

Eduardo: That is an interesting question. I think, it depends if you manage to build a model of the Stradivarius using the  physical modeling. It's very, it's a practice that has been adopted by musicians a lot, to develop physical models of such and such instrument and so on. If you can put the detailed parameters of the acoustical behavior of a Stradivarius, and then plug in an old, you know, cheap violin in it and make the system change the sound on the spot. I think theoretically it's possible.

But I don't think it's a computing problem. I think it's a modeling problem. It is, as long as, if you got the model and the model is realistic enough to give you that very fine, detailed results that differentiates, say cheap violin from a Stradivarius, then you have to find a way to implement it. And probably then because quantum computers may be faster, or more reliable or somehow more suitable for this kind of modeling, then perhaps yes.

Yuval: I'll settle for a Guarneri as well, doesn't have to be a Stradivarius.

Eduardo: Yeah. I think you'd better have the Stradivarius yourself.

Yuval: I wish I could afford It.

Eduardo: It's cheaper than a quantum computer.

Yuval: Well for now, but we'll see.

Eduardo: For now, yeah. Yes, of course.

Yuval: So Eduardo, how can people get in touch with you to learn more about your work?

Eduardo: Well, you can search me on Google, probably you'll find me there but my email is eduardo.miranda@plymouth.ac.uk. That's my university address. And it's very likely that I will read it.

Yuval: Eduardo, thank you so much for joining me today.

Eduardo: Thank you, it was a pleasure to talk to you. Thank you.