Last week I was in the studio with the BBC Philharmonic, conductor Jack Sheen and soloist Zubin Kanga, working on a recording and performance of my new piece “TECHNO-UTOPIA”. It was such a fascinating and fantastic time, I wanted to write down some of my immediate thoughts and responses specifically regarding the technology. I’ll probably do another post about the piece itself when it is broadcast.

“TECHNO-UTOPIA” is a 30 minute concerto for soloist performing on piano, synthesizers and embedded AI instruments. I had in my head ideas about magic, memory, humanity and ruthless algorithms when composing it. While they take up a relatively short amount of “airtime”, it’s those embedded AI instruments that often initially capture peoples’ attention when learning about the project.

What do I mean by embedded AI instrument? Well, simply, it means a new musical instrument that uses AI to allow the performer to express themselves in new, or different, ways. In this case, I’m using an instrument called a stacco (pictured), which builds a kind of sonic “map” based on whatever dataset of sounds you give it. In this case, it was trained on the BBC Philharmonic’s archive of radio broadcasts, so it creates this gloopy, soupy sonic landscape of uncanny orchestral sounds. This map is navigated by a performer controlling four magnets. It’s a really hands-on and intuitive instrument, and was developed by my colleague Nicola Privato at the Intelligent Instruments Lab, along with his colleague Giacomo Lepri.

Another embedded AI instrument in the piece is a concatenative synthesizer, controlled using a ROLI Seaboard. A ROLI seaboard is a kind of keyboard that tracks not only which notes you are playing, but also how hard you are pressing the keys, microtones between the notes, and so on. The concatenative synthesizer it controls is also trained on the BBC Phil archive. In short, this kind of synthesis takes a source sound and tries to recreate that source sound by very quickly scanning the archive for fragments that match the source’s pitch. So, you play a middle C on the ROLI, and the concatenative synthesizer mosaics together every middle C it can find. The soloist (Zubin) can control the mix between the source sound and the mosaic/concatenative synth sound by pressing down harder or softer on the ROLI keys.

You can hear some examples here:

So both of these embedded AI instruments are trained on just the BBC Phil’s archive, and nothing else. I’ve written before on Music Patron about how I believe in ethical creative AI, that doesn’t scrape the internet for everyone’s music, but rather responds just to data you have permission to use, provided by project partners.

What struck me about the rehearsal process was how open so many of the orchestra members were to these ideas. Of course, musicians are often an open-minded bunch, but nonetheless AI technology is a controversial and difficult subject. To be honest, I expected to have to convince them that what I was doing was interesting and useful, and not boring and unethical. But this wasn’t the case. Everyone, frankly, was very interested, very nice to me, and many members of the orchestra were keen to have a go on the instruments during breaks.

Perhaps this was because I gave a 5 minute intro to the piece before the first rehearsal, saying where I was coming from. Or perhaps it’s because the AI, in this case, is clearly embedded inside an instrument – ie, something a human uses to express themselves, that doesn’t replace anyone. It also doesn’t sound like anything you’ve heard before. Or perhaps it’s because the work is about so much more than these two AI instruments, so it was clear the music wasn’t about showcasing some technological novelty but rather about, well, music.

All I know is, we had a great time and I can’t wait to share the broadcast with everyone when we have finished the final mix! Stay tuned…