All the noise around music and AI is around Generative AI, for good reason, but it is easy to overlook the tremendous impact of Artificial Intelligence as a whole on the music industry. This article will attempt to «look under the hood» of the various types of AI that are relevant, and how they could impact the music ecosystem through all steps of the value chain.
There are several ways of classifying types of Artificial Intelligence, but for our purposes, let’s focus on categorization of type based on how (well) AI mimics human capabilities and functionality related to the music industry.
Capable AI
Categorizing AI after capabilities:
- Narrow AI
- AI designed to perform a specific, predefined task within a limited domain.
- It cannot perform tasks outside its programming or adapt to new domains without significant reprogramming.
Relevant use case: Recommendations systems like Spotify and Netflix.
What else?
Narrow-AI, also referred to as «weak-AI» excels at performing specific tasks and use Machine Learning (ML) models like supervised learning (e.g., email spam detection), unsupervised learning (e.g., clustering) and reinforcement learning (e.g., game-playing bots) to become well-trained in making decisions and predictions.
These capabilities are already in play in the music industry, the recommendation engines mentioned, but there’s a wide array of use cases for this type:
Trend Forecasting/Promotion. Using predictive AI models to analyse music trends or sounds to predict what can become popular or to identify new promotional opportunities for artists (our «bread and butter» at notefornote).
Music Discovery – using AI «similarity search» to help a user find new music.
Production and Sound Design – using AI for the mixing and mastering processes for EQ, compression etc.
Composition and Arrangement – using AI to suggest instrumentation, harmonies and chord progressions etc.
How Does Narrow-AI and Generative-AI differ?
To rewind a bit, so Narrow AI is task-oriented and cannot perform tasks outside its dedicated domain (without a serious «rewiring»).
Generative AI is a subset of AI focused on creating new content (e.g., text, images, audio, video) using existing data as input. It uses machine learning models, especially Generative Adversarial Networks (GANs) or transformers like GPT, to produce novel outputs. Furthermore, it is creative-oriented generating original content, mimicking human creativity, or augments creative processes.
Examples include ChatGPT (text generation), DALL·E (image creation), and music composition tools (AIVA). Narrow-AI cannot «create» or innovate whereas Gen-AI is capable of answering open-ended prompts, generating visuals and sounds, or even simulating conversations. It is also, «by nature», more flexible; and can adapt to new data and contexts within its creative capabilities.
Rewinding all the way to the intro of the article, I mentioned all «the noise» generated by Gen-AI, and anyone remotely connected to the music industry as a player or consumer cannot have missed it. There is great concern about rights management, revenue distribution, and even the concerning the status of creative identity itself.
Unfortunately, many tech-founders and investors seem to believe that a creative work, generated by an AI solution trained with ML models on historical data constitutes a completely «new creation» , and that the data sourcing is «fair game».
Using unlicensed music to train AI led to the major record companies suing music gen companies Suno and Udio. They let the users generate «fully formed» songs from prompts. The companies claim that these songs constitute «intermediate copies», and fall under «fair use». This argument falls more under «far-fetched», and these companies seem to have taken Facebook’s (Meta) old credo to heart: «move fast and break things»
Speak of the devil, Meta secretly used a piracy database Library Genesis (LibGen), a notorious so-called shadow library of pirated books that originated in Russia, to train its generative AI language models. Two novelists and renowned comedian Sarah Silverman, filed a class-action lawsuit against Meta in July 2023, alleging the tech giant trained its language models using their copyrighted work without permission. The case is ongoing, but the damage to Meta’s reputation is irreversible.
A Key Change – What’s next?
Agentic AI, characterized by its autonomy and capacity for self-directed action, can revolutionize the music industry by introducing applications that go beyond traditional AI tools. Agentic AI is a stepping-stone to General AI which we’ll get back to. The main characteristics include an ability to initiate actions and achieve objective with minimal human intervention and continuous learning from its environment. Today it is implemented in f.ex. AI-powered trading systems, proactive chatbots and self-driving cars.
Next Level Personalization
AI is certainly in use regarding recommendation algorithms etc., and Agent AI could push it up a notch by generating hyper-personalized playlists that evolve based on user preferences, behaviors, and contextual factors like time of day or activity. Further areas could include DJ’ing: creating and performing live mixes tailored to audience reactions, mood, and event type using real-time analytics. And, virtual concerts could be orchestrated with dynamically selected set-lists and visual effects tailored to audience feedback in real-time.
Marketing, Promotions and Fan Interaction
Fan Experiences could be both more personalised and FANtastic. Immersive, interactive content, such as AI-generated music videos or virtual interactions with Agentic AI versions of artists could offer new modes of interaction. Agentic AI can effectuate a constant online presence by autonomously managing artist social media accounts, creating posts, and responding to fans.
Talent Search
Talent Scouting could be auto-piloted by analysing global streaming data, SoMe trends, and emerging cultural patterns to discover unsigned artists with high potential.
Artists and producers and other collaborators could be matched based on complementary styles and potential for innovation, fostering unique collaborations. Sort of like a dating-app for artists driven on AI matching algorithms.
Composition & Production
So, how about real creative endeavour? Having your very own «AI co-songwriter» will appeal to many, and getting out of a songwriting «funk» is getting easier than ever. Virtual co-writers autonomously suggests lyrics, melodies, or harmonies in real-time, and Agentic AI could empower solutions that would profoundly affect composition and production.
Functionality applied to sound design could develop unique sounds or instruments tailored to an artist’s vision, using a blend of procedural generation and creative analysis. There is also an interesting potential for Adaptive Music Creation, generating dynamic, interactive soundtracks for video games or VR experiences that adjust based on user actions and environments.
Tour Planning
Using AI to analyze data and predict outcomes is one of the more impactful functionalities , and we believe it’ll be more and more central to all promotional activities in the music industry. One such example is AI-Driven Tour Management, automating tour planning, venue selection, and ticket pricing by analyzing fan demographics, logistical factors and historical performance data.
The Next Beat: Future Vibes
Our first refrain was centered around Agentic AI, which is a theme that builds into a «symphony» General AI (Strong AI). This is a hypothetical AI with human-like cognitive abilities across diverse tasks and domains without constraints. In this type of setting, the «tools of the trade» would most likely not change radically, but greatly evolve.
However, there would be several fundamental questions regarding creator identity, copyright law and the overall, global business structure. One, thing is for sure it’ll «shake your foundations» harder than any AC/DC track!
Artificial General Intelligence (AGI) would encompass a superhuman «Creativity» and could create entirely new musical genres, sounds, and compositional techniques that push the boundaries of human imagination. It might blend global cultural influences, integrate non-musical data, or use scientific principles to inspire innovation. At this stage, there’s more disharmony emanating from copyright and legal entities as well.
Copyright and Legal Frameworks
Despite all the legal ruckus around AI generated music, I think that the fundamental legal argument is pretty straight-forward, the source material for training AI is not «free use» and this must be recognized by all entities. We simply yet again see technology moving faster than legislation and the latter will catch-up.
With AGI-generated music, questions about ownership, royalties, and creative rights could become more complex. Should AGI creations belong to their developers, the artists using the AI, or be considered public domain? Esp. since AGI would be capable of generating new sounds without the Machine Learning process based on existing song material.
On the flip side of this coin, AGI could also be used for Automated Copyright Enforcement, detecting and addressing copyright infringements instantly, ensuring fair use and attribution for artists.
The Artist as The Man Machine
«Man Machine, pseudo human being
Man Machine, super human being
The man machine, machine…» (Kraftwerk)
Musicians could use AGI tools to enhance their abilities, such as perfecting their technique, exploring new creative directions, or get out of creative dead zones. The lines between human and AI based composition and performance will get blurrier and blurrier.
I have previously referred to the shifting of identity from composer to curator that is at play when the compositional process consists more of prompts than of piano playing. These AI tools could liberate the composer-curator from technical and knowledge constraints, but could impose new limits to creativity when you are denied the struggle to find the right sounds.
Back in 1985 a hit in social studies was released, Donna Haraway’s «A Cyborg Manifesto» was published. Her insistence of seeing the boarders between humans and machines as rigid rings very true in the context of creativity and AI. Using AI tools can be framed as a practice of embracing hybridity, dissolving boundaries, and challenging traditional notions of authorship. Like Haraway’s cyborg, this engagement with AI reflects a posthumanist ethos where humans and technology co-create new possibilities, but also demands critical awareness of the systems in which these tools operate.
AGI could also act as self-sufficient virtual artists, writing, performing, and even interacting with fans. These virtual entities could compete with or complement human artists, potentially becoming chart-toppers. This would be the ultimate wet dream of tech corporations, generating entertainment revenue without having to pay humans.
Ultimately, finding creative pathways, commercial structures and legal frameworks is not likely to be all sweet harmony. The Man-Machine construct raises issues concerning authenticity that impact the creator and listeners alike. Will there be more cultural homogenization or diversification?
AGI could either homogenize music by optimizing for mass appeal or diversify it by exploring niche styles and blending cultural elements in innovative ways. And, will every single player meed to be an «AI virtuoso» to have a job in the industry? AGI will take on roles traditionally filled by musicians, producers, and industry professionals, it could disrupt careers, creating a need for new roles or ways to adapt.
Super Tra(m)ps?
The final (hypothetical) stage of AI, is Super-AI. This is when AI surpasses human intelligence in all aspects and is self-aware. It’s fair to say that Super-AI is more divisive than Nickleback, and discussions whether it could exist are just as heated as discussions whether talent actually exists in said band! As, unlike Nickleback, it is self-aware, by the time we get to this particular stage, our AI overlords have most likely decided that all human expression is purposeless anyway
The Stats Rock
- The global generative AI in music market was valued at approx. $440 million in 2023.
- The market is projected to grow significantly, with some forecasts predicting it will reach $2.8 billion by 2030, representing a compound annual growth rate (CAGR) of over 30% from 2024 to 2030. (Source: Grand View Research)
- By 2033, the broader AI in the music industry, including applications beyond generative AI, is projected to reach a staggering $38.7 billion. (Source Scoop Market)
