“Artificial intelligence (AI) is so prevalent a topic today that, like irony, it’s a word or acronym that most seem to use without fully understanding its meaning. Broadly, AI is technology or technologies that allow machines to simulate human intelligence and problem-solving capabilities. When implemented in combination with other technologies such as robotics and geolocation, AI produces more accurate maps, autonomous vehicles and the generative AI tools that one finds in the news daily. Open AI’s ChatGPT is a widely-used example of a virtual assistant. AI has powered Apple’s Siri and Amazon’s Alexa for years now. From voice recognition, to web search, to customer service to weather forecasting; AI is a part of each of our days.
The above is only the most basic definition of something that will surely shape the rest of our lives as well as those of future generations. Anything written about AI now will surely be made redundant almost instantly. Despite that fact, we feel it important to state a position on both the dangers and opportunities AI poses and provides for our business today.
One immediate threat that thousands of artists and labels have voiced is AI’s ability to generate reliably realistic music on spec. Though the cloning of Taylor Swift is no doubt of great concern to that artist and her fans, one element that those of us in classical music see is in the area of licensing, especially sync. PENTATONE, as well as other classical labels and sync representatives, are regularly asked to find a work from the vast repertoire that would suit a movie scene, a student film, a commercial video, etc. If our team presents what we feel are strong options from our catalog, we negotiate a fee for the commercial use of the sound recording if one of our choices is used. If a track is not in the public domain, then the music supervisor/client must also secure that right and pay publishing. Today, AI can reproduce reasonably accurate musical facsimiles to suit the tastes of the client. And of course this can be done at no cost and requires very little time.
The consumer price increases recently implemented by many of the streaming services are a laudable step, though a marginal one. As long as pro-rata share continues to be the dominant income-sharing model for the streaming industry, an ever-widening pool means that more tracks are competing for the attention of a still-widening group of listeners. Many of those listeners are in emerging markets. Now imagine that some of the ca. 100,000 tracks added daily are AI-generated and royalty-free. The platforms can boast of delivering value to their shareholders and tout to their customers that nearly all of the world’s recorded music is available. Even without the technologies that AI supplies, there is an enormous market of sound-alike songs on most of the platforms now, many of them generating millions of monthly streams. AI-generated songs and tracks will only hasten this trend and diminish the revenue pool for rights-bearing music. $12.5 billion is lost to piracy every year in the US alone.
I do however think it would be foolish to fear AI. As a composer, we in the field recognize that AI is already a compositional tool used now like desktop notation programs were a disruptive yet life-changing tool thirty years ago, and the pencil and eraser long before. Technology has always powered human expression; AI is no different.
Labels and rightsholders should look at AI in its broadest sense as a tool that will improve a great many processes that are complicated today. Perhaps the most salient issue is that of the retrieval of musical information, where AI can play the role of an uber-librarian of recorded sound. AI provides us with the opportunity of categorizing every track in our catalog to best identify tempi, mood, length, the absence or presence of publishing, the expiration of copyrights and surely many things not yet imagined. AI can help with the generation of images, static video, animations, and any number of tools that are often demanded by the platforms and which direct customers to discover our work.
Responsible AI is a curious term that implies that the use of artificial intelligence is likely to lead to irresponsibility at best, and chaos at worst. It refers to AI guided by sound ethical and legal principles. While I appreciate the desire of the tech platforms to do what they can to guarantee this usage, I do not believe it to be possible as long as there is a single bad actor willing to mimic, copy or steal the intellectual property of another. There are now more than two dozen AI music generators, many at no cost to the user, that can generate a reasonable clone of a Vivaldi concerto, for instance. While the PROs, publishers, record labels and – most importantly – artists themselves demand the responsible use of AI, it pays to remember that we have seen similar inflection points during the early days of streaming, of file-sharing beforehand, and through various format changes. But this is wholly different.
The RIAA, in conjunction with the three major labels, have brought a lawsuit against the AI startups Udio and Suno who have defended their use of unlicensed music, citing “fair use”, which most certainly does not apply in the cases here. These are straightforward cases of copyright infringement. We can expect similar suits to brought in the coming months and years and the particularities of each litigated for some time. The music industry is only in the early stages of these challenges and “will continue to press policymakers to create new laws making it illegal to train AI models on copyrighted material without licensing deals” (MusicAlly).
Though action on AI has been slow in the United States and elsewhere, the European Union, authors of the EU AI Act, offers the first comprehensive laws aimed at assigning levels of risk to AI use. The Act is in many ways an updated supplement to the GDPR legislation of 2018. PENTATONE is a supporter of The Principles for Music Creation with AI, which “establishes a guideline for the responsible use of Artificial Intelligence in music. By adopting these principles, companies and organizations commit to protecting the vital contributions of human creativity and to the responsible development and application of artificial intelligence for music creation.” Among classical labels, we are also proud to be one of the first signatories of the AI for Music initiative.
Classical music is a great many things but one thing it offers that no other genre can: a millennia of history. What most would agree to be classical music has been around for ten times as long as jazz, fifteen times longer than rock, and twenty times longer than hip-hop. Fans and listeners with deep appreciation know that a Palestrina motet, a Wagner opera, or an Anna Clyne concerto have no equivalent in any other genre. We at PENTATONE believe in providing the highest quality listening experience in terms of artistry and sound, and all within the broad and undefinable genre of classical music. That breadth means that we must strive to understand, share and create demand for all of our recordings, and to meet the challenges that recorded classical music faces. Though this is increasingly difficult in a world with a very short attention span, I for one think that that complaint has been whispered for centuries. Despite it all the music appears to live on and thrive. Among the PENTATONE team are singers, pianists, early music specialists, composers and musicologists. Each is a valued member of our family; they make it possible for you to discover and experience the very best in recorded classical music.
We at PENTATONE can offer no quick solution to the threat of AI, nor a wanton embracing of AI technologies to assist us in our daily tasks without a thorough and growing understanding. But we recommit ourselves to our organizing principles in delivering the greatest range of world-class artists, presented in premium-quality productions, each captured in exceptional sound. We invite you to take the journey with us.”
Sean Hickey, PENTATONE’s Managing Director
July 17, 2024