Music, Health & AI: Making Unwind.ai with Marconi Union

The idea of generative music how it should be...using it relate to people, our own biology to either create or tailor music to themselves. It finally seems that generative music ... has more purpose
— Duncan Meadows of Marconi Union, speaking about Unwind.ai

For our first episode of the Sync Project Podcast, Marko Ahtisaari sat down with Richard Talbot and Duncan Meadows from the critically-acclaimed band Marconi Union, who helped us make Unwind.ai, a personalized music experience designed to help you relax before sleep. 

Richard and Duncan talked about the creative process behind making “Adrift”, the first composition available through Unwind.ai, their view on modern music technology and whether artificial intelligence and generative techniques in music are a tool or a threat to musicians. 

The interview also revealed the band’s beginnings in a Manchester record store, what it was like working with other influential artists making music for sleep like Max Richter and their plans for the future -- and the role generative music like Unwind.ai may have in their next releases. 

For a full tracklist and more information on Marconi Union, check our the podcast  and don't forget to follow Sync Project on Soundcloud.

Read more for a full transcript of the interview:

(Intro)

Marko Ahtisaari: Welcome, we have Duncan Meadows and Richard Talbot from the incredible Marconi Union, a group we’ve been very, very happy to collaborate with at Sync Project on a project called http://unwind.ai, where we built on a piece called “Adrift” that we made together through a series of conversations and now is online. The music we’ve co-created with machines is now being being by thousands of people to sleep to, and relax before sleep, at http://unwind.ai. We’re really happy to collaborate with you and have been personally following you for some time. The first question to ask is, not that our collaboration is live online, are you curious about what is going inside people’s heads and bodies a they listen to “Adrift” as it’s being manipulated by Sync Project’s machines?

(1:30) Duncan Meadows: I’ve been curious for awhile, ever since we created the “Weightless” project a number of years ago, partly because a music therapist with whom we consulted told us about a process called entrainment, where the body’s biological processes can be begin to synchronize to an external audio source. So, that in itself is quite interesting, being involved with Sync Project and contributing to the goal of understanding how music can help people was one of the reasons was one of the reasons we wanted to get involved. And if the scientists who are involved can find out more about what’s happening inside the brain while they listen to music, that’s fantastic. We are interested, and as a science, it feels like we are just at the start and in the decades to follow, hopefully we can learn more.

MA: Could you tell a little bit to the audience that hasn’t heard “Weightless”: 20 million streams on Spotify and a quite a bit on Youtube, 2011 you released that

(2:50) DM: Looking back no-one involved had an idea of what would happen with “Weightless”, but personally, the gratifying thing for me is that we heard about people (and they contact us) about how Weightless had helped them overcome and manage different health problems. Although we’re very keen to stress that for us “Weightless” is a piece of music, the fact that people are taking these things away is incredible. This was another reason we felt Sync Project was giving us another opportunity to try and help more people. That was never our goal as musicians, but if we can help then fantastic! We know instinctively that musc is a form of therapy, but the question is how is it doing that? It was interesting, “Weightless”, in one way, sent us down a road we never thought we’d go down, and it’s been quite a ride.

MA: One of the things you told when we starting working http://unwind.ai and your piece of music for that, was that people were making repeat loops of Weightless on Youtube, and releasing them, till you made an official long-form version of it.

Richard Talbot: … We had quite a few requests for longer versions of it, and people were making their own, but decide to scale on that idea and make a 10 hour version ourselves. 

MA: For the audience that’s listening, worked together on a new track called “Adrift” we took the stems and then our computational and generative music engine manipulates those based on physiological input. Maybe you could both say a little about the process behind that, the good, the bad and the ugly...

Try unwind.ai, personalized music for relaxation designed to help relax before sleep, featuring exclusive music from Marconi Union

Try unwind.ai, personalized music for relaxation designed to help relax before sleep, featuring exclusive music from Marconi Union

(5:03) DM: I think for me firstly was the opportunity to work in a generative sense and discover what the limitations were, but was possible, that was interesting. The software that we used I still like to dig further into it, I think there’s still a lot more to be accomplished...We tried not to approach it as a scientific thing, we always sort of think of it as music first and then, how could it be incorporated into a machine… It was fun watching it unfold, and develop over the months. I’m proud of what we’ve done together, it’s very good.   

(6:00) RT: I think the only thing I really want to say is that the idea of generative music is something that’s appealed to us for a long time now, and we have used elements of it in in making our own music. We’ve used various well-known pieces of software like Noatikl, but also setup systems in Ableton and Logic. [Unwind.ai] was an opportunity to do that on much bigger stage, if you like, and push that right to the front of the whole process of making music. That was pretty exciting and we will definitely be exploring those ideas further.

MA: You’re as a group, critically acclaimed and followed, and when Brian Eno books a festival, he books you first and in a way he’s the person whose coined the term “generative music” (even though it’s been used in computer music for some time). With our collaboration, what we’re essentially doing is generative music with physiological input. How do you see that exploration proceeding in generative music? Maybe one angle to that is, more artificial intelligence and machine learning in music as whole: do you see it as a complement to your work, a tool, a threat?

(7:20) RT: Up until now, speaking personally from my experience, it’s just been a tool for generating ideas or basic sketches of things and then we intervene quite heavily and work with whatever we have. It tended to be ideas that were quite stochastic, they’ve quite randomized forms, with a bit more depth with Noatikl and so on. It’s something we’d like to explore further… 

(8:10) DM: Yeah, I hope musicians aren’t replaced by algorithms in the future, so perhaps I see it as a threat, but I’m from more of a traditional musician background. The only thing I’d really say about it is that I sometimes try to imagine myself as producing generative music, outside of the usual timing parameters. Musicians nowadays write with grids in front of their faces, it’s all in the screen. Some musicians, and I fall into the same trap sometimes, you’re obsessed with bar lines and grids and everything looking good  and I think it’s important to remember that you actually music is … humane, it’s about being human. I sometimes I feel that I’m not sure what generative music is about really. I understand it’s about producing patterns and randomness, but I don’t have as much interest in it as certainly Richard does. 

MA: And certainly “Adrift”, which was the track that was the input into Unwind.ai, doesn’t sound strongly barred. It sounds very organic in the way that it unfolds.

(9:33) RT: That’s a bit of a thing with a lot of our music, to be honest, we’re a very textural kind of band. We don’t to tend to reach to conclusions. It almost feels as if it could be an excerpt from a larger piece. That is something I think we are quite attracted to and it’s partly it comes from the way we work: we accumulate ideas and pass them on, then this process of addition and subtraction, and that’s passed on again, and eventually we end up with some music.  

(10:15) DM: I think Rudy Van Gelder, and I’ll paraphrase him, he spoke about how he wanted the studio and the technology to sound human. I think we try and do that. I don’t like music where you can hear the technology so much, I think it needs to sound as if it’s come from a human place. I think what Richard said about our music sounding like it’s part of a process, I think sometimes it does sounds like you’ve dipped into a moment, there’s a past and a future to a track that you’ll never hear, and I like that about our music. With Adrift, the sense of it have no particular timing, that is a bit of a technique actually. There is music theory behind that, not that we sat down and thought about it too much, but that sense of it floating free, there is thought that went into that, it’s not accidental. Actually, with Adrift, we’ve learned some things that we’re going to use in the future certainly, that may appear in one or two albums time. We’re keen to develop those ideas, which is another reason for being grateful for having been involved in it. Anything that provides inspiration and enables you to go down a new path is something we try and embrace.

MA: Happy to hear that. I think our approach in terms of machine learning artificial intelligence used to create music, our approach is to do together with artists like yourselves, and then combine human physiology into that equation. That’s kind of the triangle in which we operate and we think people are also more intrigued by that, rather than purely robot-made music. I think unwind.ai sounds authentic and human. In fact, it’s the most human music in some sense because it’s starting from your physiology so you have certain associations with it

(12:29) DM: Exactly, and what I was saying about being suspicious of generative music, I think things like Adrift [Marconi Union’s track in Unwind.ai] and other things Sync Project will do in the future, I think that’s using the idea of generative music, for me, how it should be. Using it to relate to people, and actually using, as you say, our own biology to either create or tailor the music to themselves. To me, it finally seems that generative music, and the software and the program behind it, now has a use to me that makes sense. Finally, to me, it feels it has more purpose.

MA: Another way machines are being used today is for curation. The origin of Marconi Union, originally as far as I know, is from a record store in Manchester. I just wanted to ask Richard, Duncan you joined as a keyboard player and full member later, do you see that as a record buyer that has influenced your process still? And that act of a piece of music  being like a moment that something came before it, the soundtrack and after it, Richard, do you have any thoughts on that?

(13:52) RT: I often feel like I’m just dipping into worlds of music, you know, I might decide I want a dub record but then you get this sense of it being part of being a part of a huge body of work. Lots of different people making dub records and a lot of music I end up constantly sampling it rather than really explore it, because there’s just too much of it. And that’s quite exciting that idea, just grabbing something and listening to it, you know. I’ve got no idea what this is, where it’s from, who made it or anything about it and I can just judge it, as to how it affects me. It’s quite selfish really

MA: You’ve also done some remarkable remixes and one that struck me personally is your remix of Max Richter’s “Dream 13” from Sleep, somebody we’re working with and also much respect. Have you experienced the full 8-hours of Sleep, yourself? 

DM & RT: No, we haven’t.

MA: And thoughts on that remix? For the audience that doesn’t know, Max Richter wrote an 8 hour piece to sleep in, and several segments of this piece were remixed and Marconi Remix did a remix of the track Dream 13. We recently tracked the sleep of several sleepers at Old Billingsgate Market, where 450 people were asleep to a live performance of Max Richter’s Sleep
And thoughts on that? Where you’re remixing a piece purportedly for a purpose, it’s a “piece of repose”, as Max would put it?

(15:46) RT: I think we moved away from that. We didn’t follow that line of this is going to be about sleep and actually moved into something that’s actually quite a bit more upbeat and brighter, and present maybe

MA: The daytime version?

RT: Yes, absolutely

MA: And finally, what are you working on now? You have Ghost Stations and Ghost Stations Remixes out, Tokyo+ recently came out, all remarkable pieces of work. What are you working on now?

RT: We’re just kicking around ideas for music and we’ve no idea where they are going to go. Whether they’ll go anywhere or not. We’ve got a gig next month (Saturday July 15, Band on the Wall, Manchester) and that’s our last one for this year. We don’t rush things … we’re not known for being overactive about this stuff. I guess what happens is, it’s that process like talked about with generative music. We have get that process together about how we’re going to move on and make the next piece of music, we have to decide what we’re interested in, and how we’re going to pass ideas around between us and what those ideas are going to be. It all takes time. It’s painfully slow, but we’re happy to take time… We don’t like to put out things for the sake of it. That was a non-answer wasn’t it?

MA: I thought what you were saying is that while you talk relatively little about the process and the tools with which you make the music, it seems like you very reflexive [about how that’s done]

(17:58) RT: We don’t use anything different from anybody else, there is no secret weapon. We’re not using some sort of highly arcane software or anything like that, we just use all the same off-the-shelf stuff. It’s the process that makes comes out be Marconi Union.

MA: Before each project or each stage of work you are quite deliberate about thinking about what the process will be. That is an answer and not everyone would say that. I was speaking to another electronic musician that I think we both respect, and they did a complete reset of tools and process for the new record to kind of check the tools, that they hadn’t become slaves to the system.

(18:46) DM: You do have to reset to avoid repetition. It would be easy release an album a month I think but it needs to be hard to otherwise for me it doesn’t feel like you’ve accomplished anything. It’s a good point. Although we are using the same equipment as everyone else, there are ways of approaching that equipment and configuring things and just spending time with equipment. There’s a danger that there’s so much technology and software that you can surround yourself with too much and don’t make the best of what’s available. Personally one of my principles certainly for the next record, and recently, has been more about reduction and simplifying and for me I feel like I’ve become more creative as a result of that.