Daniel Johansson: Analysts are starting to talk about an AI bubble, similar to the dot-com bubble

Datum: 5 July 2024

Music industry researcher Daniel Johansson summarises the first months of 2024 and looks ahead: what can we expect for the rest of the year?

COLUMN

In 1999, I wrote my first tech-oriented news and analyses for the Swedish music industry. One of the very first news pieces I wrote was about the release of Napster. Since then, I have written a few thousand articles and analyses.

I look back a bit in my database and find this news item from November 1999:

“CDnow will start selling music for download this Friday. The first batch will contain 13,000 songs from artists including Beck, Frank Zappa, Fatboy Slim, Frank Sinatra, Muddy Waters, and many more. Prices will range from $1.49 to $3.49 per song, and will only be available in Liquid Audio format.”

This was four years before the iTunes Store was launched and nine years before Spotify was released. A lot has happened since then.

I have had the privilege of being part of the shift that the music industry has undergone since the late 1990s. Or rather, multiple shifts: the transition from physical media to downloads, from downloads to streaming, from music consumption on computers to primarily mobile devices, and now, the transition from “dumb” technology to intelligent technology.

Contemporary Technology
The relationship between music and technology has always been strong. Contemporary technology lays the foundation for the creation, production, performance, and consumption of music. New formats and exposure platforms introduce new ways of working and operating, and it has always been the actors who manage to adapt to the new technical frameworks that succeed the best.

The first news piece I analyzed for MI in 2024 was that Elvis Presley, starting at the end of the year, will go on tour as an AI hologram:

“The concert is called Elvis Evolution and is developed by British Layered Reality. It seems they will mix different technologies, augmented reality that requires some form of glasses, some form of multisensory technology, and new projection technology. The concert seems to be somewhat different compared to other hologram concerts, such as ABBA Voyage which primarily relies on large LED screens and advanced lighting, or Whitney Houston, Roy Orbison, Dio, etc., which have been projections of avatars on a film of transparent plastic. After the concert, attendees will be invited to an after-party. Whether AI-Elvis will participate in the after-party is not clear.”

A Paradigm Shift
A significant portion of the news so far in 2024 has, of course, been about AI. We are going through a paradigm shift that may be greater than what the entire internet meant, and probably no one really knows where it will lead. Here is a selection of what I have written on that subject since the New Year:

  • Microsoft integrates the AI platform Suno into Copilot, enabling music creation directly in Copilot.
  • The American artists’ organization SAG-AFTRA enters a collaboration agreement with Replica Studios regarding licensed AI voices.
  • TikTok introduces its new AI Song feature in beta version.
  • Universal Music Publishing and others initiate legal proceedings against the AI company Anthropic, which has been trained on lyrics represented by the publishers.
  • The Swedish industry organization KLYS launches an AI strategy designed as a tool for creators in various industries.
  • SoundCloud partners with three AI companies, Fadr, Soundful, and Voice-Swap, whose technology will be integrated into the platform.
  • The EU Parliament passes the so-called AI regulation.
  • German GEMA and French SACEM (equivalents to Sweden’s STIM) publish a study on AI and the music industry, with 15,000 songwriters and music publishers participating, showing that over half of all music creators under 35 use AI tools.
  • Alibaba demonstrates an AI model capable of generating videos of singing people based on just a single image.
  • TuneTaxi from the Netherlands generates 1 million AI songs and will launch a platform targeting the entire background music market.
  • YouTube begins requiring users who publish AI-generated content to tag it, introducing a new feature for this in Creator Studio.
  • British BPI initiates legal proceedings against Jammable (Voicify), the most common tool for so-called AI covers.
  • The state of Tennessee in the USA introduces new legislation making it illegal to use AI clones of others’ voices without prior approval.
  • Universal and Roland develop joint guidelines on responsible use in connection with AI and music creation.
  • Soundry.ai releases a plugin that enables music prompting.
  • Over 200 artists, including Billie Eilish, Jonas Brothers, Chuck D, Sam Smith, Katy Perry, Pearl Jam, and Nicki Minaj, publish an open letter urging platforms to handle AI in a way that does not violate their rights.
  • The new AI platform Udio launches, generating high-quality music and vocals, and quickly gains widespread adoption.
  • Aftonbladet tests presenting news as rap, with the help of AI.
  • Wyclef Jean, Justin Tranter, and Marc Rebillet are among the artists who have made music with Google’s new model for AI-generated music, called Music Sandbox.
  • BMAT and Voice-Swap introduce a certification for AI platforms to ensure they train their models on approved content.
  • Sony Music sends an open letter to 700 AI companies, demanding that they obtain permission to use their music for training.
  • Eleven Labs raises 800 million SEK in capital and showcases its new AI platform for prompt-based music creation.
  • Randy Travis, who lost his voice in a stroke in 2013, releases new music with the help of an AI clone.
  • Washed Out releases a music video in collaboration with Open AI and the video model Sora.
  • Suno and Udio are sued by the American industry organization RIAA.
  • Universal Music Group launches an AI plugin for its artists, allowing them to create AI models of their own voices.

Just a selection of everything that has happened regarding AI and music in recent months.

We can expect the “battle” between AI platforms and rights holders to expand even more during the rest of 2024, especially now that the first lawsuits have occurred. In many ways, the situation resembles what happened in the 2000s with piracy, when various tech companies referred to what is known in the US as “fair use.”

Now, AI platforms refer to the so-called TDM exception (text and data mining), an exception intended to allow the creation of new knowledge and research on various types of datasets. Rights holders, not just in music, argue that TDM does not apply to this kind of training of AI models, and this will likely become one of the most critical issues to resolve in the future.

I don’t think it will be resolved in 2024; we can probably expect several years of legal processes and interpretation of new legislation, such as the AI regulation. After all, much is at stake, as platforms like OpenAI’s ChatGPT, Google’s Gemini, or for that matter Midjourney, Dall-E, Stable Diffusion, Suno, Udio, and others, depend heavily on what they have been trained on.

More and more analysts are also starting to talk about an AI bubble, similar to the dot-com bubble around the turn of the millennium, or even the crypto bubble. It is likely that many of the billions currently being invested in companies that sprinkle their pitches and business plans with terms like neural networks, deep learning, reinforcement learning, and generative AI will meet the same fate as many companies founded in the late '90s.

Just like with the internet and digital media 20-25 years ago, we are in the midst of that first big hype, something that Nvidia’s stock price, among other things, testifies to. Overinflated? Yes, of course, but beyond the hype comes the next phase, when the technology settles and begins to seriously change society and the economy.

We are not quite there yet, but we are getting there. The startups that today can identify the potential “gaps” in the upcoming market, after the initial hype has subsided, are, I believe, the ones that will survive and secure an important position for the rest of the 2020s and perhaps especially in the 2030s.

More Legal Processes
In the field of music, we are only at the beginning of these changes, and for the rest of 2024, we will likely see more legal processes and an expanded “war” between AI companies and the industry. Some see this as a negative development, with copyright once again coming in to hinder technological advancement, but I see it differently.

It is precisely this balancing act between restrictive rights, ethics and morals, and rapid technological development that creates truly great products. If copyright had “died” back in the 2000s, Spotify would not have been forced to develop a product that was superior to quick fixes, and much of the Swedish music tech that has emerged over the past 15 years would likely never have seen the light of day.

So, it will be exciting to see what happens in the coming months. I will be back.

Daniel Johansson
daniel@musikindustrin.se