Steve Thompson is a musician, inspired by advancements in technology. Founder of the multi-genre ensemble 1201_Alarm, he reflects on the many misconceptions about AI in the music industry.
In an age marked by rapid technological advancements, our sector finds itself grappling with the potential and pitfalls of artificial intelligence (AI). As a professional musician, inspired by science and technology and with a profound love for both music and computers, I want to share my thoughts, dispelling some common misconceptions that cloud the discourse around AI in music and the arts.
The arts and cultural sector is no stranger to the winds of change, constantly adapting to new tools, platforms and trends. When I started composing with computers in the 1980s, I was working with little more than beeps and clicks. Fast forward to today, and the arsenal I have at my fingertips includes a full digital symphony orchestra, choir, rock band and more with a breathtaking potential for expression I never dreamt possible.
Advancements in AI technology have been nothing short of astounding. Deep learning models, neural networks and sophisticated algorithms have made it possible for machines to mimic human creativity. OpenAI's MuseNet showcases the rapid strides.
Far from infallible
But AI's capabilities are not without limitations. While it can currently reproduce existing musical genres, it struggles to create new styles. There is a marked difference between AI-generated music and compositions by human artists - the former often lacks the emotional depth and storytelling that defines human creativity. This is what must motivate musicians of the future.
Despite its remarkable progress, AI in music is far from infallible. The quality of AI-generated music hinges on the data it is trained on. It can only produce music as good as those datasets.
But is it a tool to aid musicians in their creative process, or is it a means to replace them? The answer lies in the hands of artists, producers and the broader music industry. We should be very mindful of how music is consumed and what purpose it fulfils.
Experimentation with technology
For my first album with 1201_Alarm in 2017, I used a computer to harmonise a section of music. It wasn’t AI, but a programme I wrote describing the rules of Drop 2 harmony, something any music graduate would know.
The machine produced the expected results and the band played the music. As far as I know, its origin went undetected by listeners, critics and colleagues.
For my latest album, I experimented with AI. I used the most popular and well-known tool – ChatGPT - and asked it for a sequence of chords that would create a certain emotion in the listener. I appreciate this is a huge escalation from the 2017 brief, but considering the expectations of its capabilities, I thought it reasonable.
The results were disappointing. It produced a bland set of progressions that have been used countless times before and were uninspiring and vapid. They did not make it on to the album.
AI is a tool, a collaborator, a muse, but it can never replace the innovative spirit of musicians and composers. The essence of music lies in the stories, emotions and cultural narratives that artists convey through their work.
As a musician who has witnessed the transformative power of music, from intimate pub gigs in tiny back rooms to playing to an 80,000 audience live in the Olympic Stadium in 2012, I think AI's role should be to amplify human creativity, not to diminish it.
One example of this at work is the image below. I trained an AI tool on pictures of band members and asked it to create a cover for our new album, Moonshot*. After hours of training, it produced an image of me dressed as an explorer in front of a huge moon rise.
Because of noise and data variability, AI-generated images are not always accurate
Impressive, but look closely and you’ll see I have 6 fingers on my left hand. How could something so advanced make such a simple error? Remember, AI does not know it is representing fingers. It simply sees what it has been trained on and, because of noise and variability in the data, makes an educated guess – not always getting it right.
A new wave of creativity?
AI is not a threat to musicians as long as we ensure our relevance. We have to make the case for human nuance and inflection and push the significance of human performance and creativity over technology-created music. And, importantly, we have to encourage the consumption of music for music’s sake - not just as something in the background while performing some other task.
AI has the capacity to amplify our artistry, offering fresh perspectives and innovative possibilities. Despite the challenges and misconceptions, the future of music lies in harmonious collaboration between human creativity and AI's computational prowess.
It is potentially the biggest democratisation of art since the birth of the internet. Perhaps the symphony of human and artificial creativity has just begun.
*1201_Alarm’s new album Moonshot is available on all streaming platforms now.