David Ronan joined AI Music as Head of Research in 2017, and ever since, has played a key role in some of our most exciting projects. Originating from Limerick, Ireland, he brings his vast knowledge of both music, artificial intelligence and their unique relationship to everything we do here.
We spoke with David to learn a little more about his background and help shed a light on some of the most captivating aspects of AI Music research.
Q. How did you first get into the field of artificial intelligence & music? A. Growing up, I always had an interest in AI, having watched and read my fair share of sci-fi over the years and watching the Matrix one too many times of course! My interest in music came from always being into electronic music, going to gigs and producing music with friends while growing up in Limerick. At university, I did a BSc in Mathematical Science and then an MSc in Music Technology which nicely combined my science and musical interests. This led me to do a PhD in Electronic Engineering and Computer Science at Queen Mary University of London. It’s during my PhD that I really explored how AI could be applied to music as well as intelligent audio production tools. I really love the intersection of those two topics, and I thought, why not make a living out of something you're passionate about?
Q. How do you and your team approach research at AI Music? A. Our team was born out of the same research group at Queen Mary University of London, so we’ve all been exposed to similar topics. However, we all have our own individual strengths that naturally complement each other. It’s important to strike a good balance in that respect. While it’s good to have a research background, it’s useful to have a pragmatic engineering approach as well. It’s easy to get lost in an interesting topic, but you have to be able to take a step back and ask if what you’re doing is adding value. We’re also big fans of having hackathons– pressure makes diamonds as they say!
A typical day could be reading lots of papers, coding, running experiments or having lengthy whiteboard sessions with the team. We usually find a walk to the Peace Pagoda in Battersea Park is good to get the creative juices flowing.
Q. What’s the most interesting trend in AI & Music right now? A. I think one of the most exciting things I’ve seen lately is what OpenAI are doing with variational autoencoders. They’ve achieved some very interesting results and I can see a lot of potential there. I really like how they’ve managed to retain the stylistic qualities of the music it’s trained on and made things sound coherent up to a point. There are definitely some hurdles that need to be overcome with song structure and noise, but it’s certainly an exciting leap forward in using artificial intelligence to generate music.
On a whole though, we’re witnessing a huge explosion in the amount of research applying deep learning to music. Our research team went to the ISMIR conference in Delft last year and we estimate that about 90% of the publications presented there had some element of deep learning. There has also been some really cool stuff published recently where researchers have used deep learning to do black-box modelling of audio effects like reverb and dynamic range compression.
Q. The most exciting project you’ve worked on (either at AI Music or during your own research) A. I’ve worked on two really exciting projects in the last year, one was for AI Music and the other was personal. At AI Music, we worked on a project called the Infinite Music Engine. This is a system we developed that allows us to algorithmically generate massive amounts of music at scale. By massive, I mean, if we leave the algorithm unconstrained in one pass through we are able to generate billions of metadata rich tracks. These metadata rich tracks can then be reshaped to however we want using our Remix Engine, which in itself is another multiplier. We’re currently streaming this content endlessly on YouTube, where we’ve also used AI to mix the music together – just like a DJ would.
The personal project I worked on was a multi-channel audio installation called ‘Dark Matter Radio’ that was displayed at the London Science Gallery. In the project, we took a mathematical model of the dark matter field in our galaxy and mapped this to audio synthesis parameters.
Dark matter is what makes up approximately 80% of the mass in the known universe and is made up of material that scientists can’t directly observe. Dark Matter also emits no light and emits no energy, so it was definitely a super interesting project to work on! I worked with academics from the Physics department at King’s College London and the artist Aura Satz to bring everything together. The project was published on Forbes Magazine and New Scientist magazine – so we were pretty chuffed with that!
Q. What advice would you give to anyone looking to get into this line of work? A. There’s a quote by a very famous mathematician/entrepreneur (Jim Simons) that really resonated with me recently, it’s to “be guided by beauty”. You should aim for everything you do in life to have an aesthetic component about it. Whether it be writing music, code or approaching a problem and doing it right. I think that certainly applies in this line of work since music is generally meant to be aesthetically pleasing, which goes hand in hand with finding elegant solutions to musical problems.
Q. What’s the best part about working at AI Music? A. The best thing about working at AI. Music other than the great people I get to interact with is the diverse amount of challenges we get to tackle on a day to day basis. Since my time at AI Music, we’ve focused on many different aspects of music technology. We’ve done things like digital audio effect design, automatic music remixing, algorithmic music generation, audio source separation and dealt with a whole array of music information retrieval problems.
In the last year, we’ve also started looking at how we can start using listener context and how it can affect the music you’re listening to. This has been really exciting as we’ve been working with different environmental inputs such as time of day, location, heart rate and mapping these to musical parameters. As a result, we’ve observed some really interesting feedback loops within the musical journey of a listener.
Q. What’s your favourite song of all time? A. That’s a tough question… as I’m more of an album person. I’ll give you three of my favourites in no particular order as I couldn’t possibly pick one.
Aphex Twin - Ageispolis: I just love the groove and melody of this as well the distortion on the drums. I’ve always been a huge fan of Richard’s work, so I have many more favourites by my fellow Limerick man. Selected Ambient Works 85-92 has always been on rotation.
Boards of Canada - Everything You Do is a Balloon: I have always enjoyed the nostalgic sound and natural feel of Boards of Canada’s music. I remember getting their album ‘Geogaddi’ on MiniDisc just before I moved to New York many years ago, so I’ve always got fond memories when listening to them.
The Beatles - Tomorrow Never Knows: One of the Beatles more psychedelic offerings, which has been a huge influence on a number of other artists I really like. I’ve always been really into The Chemical Brothers and I think this song, in particular, has influenced a lot of their work. All that said, The Beatles are timeless and this stand out track is on Revolver which is probably my favourite album by them (a tough call for sure).
Q. What song do you have on repeat right now? A. No song in particular, but have gotten really back into Miles Davis lately and have been listening to his album ‘Bitches Brew’ a lot.
Q. What item would you take with you to a desert island? A. If I didn’t have to worry about survival/electricity I’d probably bring my Elektron Analog 4 synthesizer. I guess you’d normally take a book in these situations as well, so it would have to be “The Elegant Universe” by Brian Greene for me. Both of those items have endless possibilities! To keep up to date with all the goings-on at AI Music and meet more of our team, sign-up to our monthly newsletter!