Music tech: who will be in charge of AI?
It's tough to keep up with the flurry of news surrounding AI, with product launches and other "revolutionary" announcements coming thick and fast… Producers, artists, composers, labels, data analysts, music tech service developers: AI is impacting every profession and permeating artistic creation and operational processes, shaking up the industry as a whole. Water & Music just held a timely webinar to take stock of the situation.
-
-
First, an observation: the market is maturing, with 25% of music creators having already used AI tools (mostly for stem separation or mastering, rather than creating a complete track). 40% of industry professionals use AI in their data processing. However, on the investment side, there has been a significant slowdown: due to ongoing lawsuits and persistent legal limitations.
Three AI trends in music tech
Faced with the scarcity and cost of "real" data, AI models are increasingly being trained on synthetic data. This eliminates licensing costs and negotiations with rights holders. Feeding the AI with its own outputs seems like a risky idea at first glance, given that AI degenerates generation after generation. However, researchers are exploring this avenue for generating plugins, instrumental variations, and even data tagging. A hybrid approach mixing "real" and synthetic data appears to be the most optimal for training models. It should be noted that the audio quality generated by AI is still insufficient to allow it to "self-feed."
– The fine-grained control of AI models is progressing, and users can increasingly guide and modify AI outputs (notably by modifying a render through prompt iteration, without having to regenerate from scratch with a new, more refined prompt). According to Water & Music, this increasing granularity is eroding the traditional advantage of composers and producers, giving greater prominence to artistic direction at the expense of technical expertise. This trend is similar to Rick Rubin's "magic formula," which he has already discussed.
– Finally, the detection of AI content marks a decisive turning point, directly linked to the protection of copyright: streaming platforms are tracking artificial content, services are offering certifications for data training models, detectors of musical deepfakes, watermarking of content… with problems for each approach.
In summary, power dynamics are emerging among a multitude of players, and it remains difficult to predict who will control the AI landscape in the near future. Water & Music concludes with a vision that is sure to provoke a reaction: artists will likely need to adopt a more flexible approach to intellectual property in order to enhance their reach, both artistically and operationally.