Discover more from Network Notes
🔵 YouTube and Stability AI: a tale of two approaches to copyright
Is this also moving too fast for everyone now? I suspect it might be.
Hi there -
AI is now casting such a long shadow over everything in the music industry that only one or two of the stories below do not relate to it in some way.
It is a real tale of contrasts too, so let's take a look.
Perhaps the biggest news story is that of YouTube's new AI modelling via its Lyria LLM, which will (eventually) enable users on the platform to create music by just humming parts in. For example, you can sing a melody and have Lyria turn it into a bassline or guitar riff.
The really interesting part however is that YouTube has teamed up with select artists like Charli XCX, Sia and Troye Sivan to clone their voices, such that they can perform within your song.
What's clear is that YouTube is very much aligning itself with the ethical side of AI in this instance; a world where artists allow their voices to be cloned and receive due payment accordingly.
This isn't a shock - it wasn't long ago that Universal and AI made their big announcement about collaborations in the AI space - though perhaps a notable detail is which labels the artists belong to.
While Warner hasn't been quite so vocal about any collaborations with YouTube, with former YouTube COO Robert Kyncl at the helm it is perhaps little wonder that the company is now supporting the new Lyria AI. Sony, on the other hand, appears curiously silent on the matter.
All of this stands in stark contrast to one of the other stories, that of Stability AI's Ed Newton-Rex stepping down from his role at the company citing concerns over its approach to copyright.
In short, Newton-Rex fundamentally opposes the company's stance that it is essentially acceptable to train LLMs on copyrighted material, a view that stands it in the opposite corner to YouTube.
Doubtless some heavy lawsuit action is coming, and these in turn will dictate how the copyright situation is resolved. I wouldn't be at all surprised if we see a few companies sued out of existence, as has happened a fair few times in other areas before now.
Spare a thought though for the work of Holly Herndon and Mat Dryhurst. If you read the New Yorker profile below you'll see that the pair are very much operating in a more pragmatic and positive space relative to protecting artists whilst also opening up opportunities for them. This is a critical area, one of few sitting free of Big Tech involvement, which I'd argue makes it a crucial undertaking, particularly in the indie space.
Chatting with a colleague earlier, we agreed that things are almost moving too quickly at present. OpenAI has already unveiled GPT5 when 4 is barely out of the starting blocks. Law is a slow, laborious thing to evolve, and right now it is painfully behind.
Some tools feel exciting, such as Meta’s new Emu system (see below). However, every new iteration brings a sense that all this has consequences. Yes, Emu’s video editing is incredible, but it isn’t a welcome development if you’re a video editor by profession. Likewise developers whose expertise is being slowly eroded by the likes of Microsoft’s Copilot AI.
My point is not to spread fear; I think plenty of people are doing that already. However the speed at which this is evolving, and where you sit within that, are something everyone should be all too aware of. While I may be belabouring the point, I still encourage everyone to engage with tools like Chat GPT+ to gain a deeper understanding.
I’ve always held a view that we shouldn’t fear something if we can stay ahead of it. Taking my own company Motive Unknown as a case in point, we 100% accept that AI will replace all manner of processes within the business. We are working hard to stay ahead of those developments, and understand the potential (and potential impact) of AI as it keeps evolving.
However, staying ahead of this all is getting harder and harder, hence my wish that it would slow down a little. These calls have already been made, but as yet no government has stepped in to actually put the brakes on. I hope that might happen soon though, as this is all progressing at a pace that one would struggle to keep up with even if you only focused on this and nothing else for your day job - and god knows for most of us this is far from the only focus we have.
Have a great weekend,
🎶 listening to “Black” by Kevin Richard Martin. Better known as The Bug, the KRM guise is Martin’s exploration of more beatless, spatial atmospheres. Calling this ambient would be a mistake; it is more dense than that, crackling with noise and tension. This particular album has a whole connection to Amy Winehouse too, but rather than paste that all here I’d urge you to go and read Kevin’s words over on Bandcamp. A wonderful listen though - something you can really immerse yourself in. Great headphones music.
📖 reading “The Blue Moment” by Richard Williams, a study of Miles Davis’ legendary Kind of Blue, detailing not just the genesis and recording of the album, but the ripple effect it went on to have through the outsider music diaspora, including the likes of Steve Reich, John Cale & The Velvet Underground, Brian Eno and more. Highly recommended.
From next week Network Notes will move back to its Mon/Weds/Fri schedule. Writing articles 3 days on the bounce whilst trying to run a company was killing me 😅
The Network Notes music league is still open if anyone wants to join. So far it’s certainly proving hotly contested! It’s free to join and you’re not handicapped by joining after the rounds have started. In fact given you can get minus points here, you might actually start halfway up the league on zero 😆
- continues to grow, so thank you for everyone showing support! Next post will go out on Sunday as usual.
Stories from the Music Industry:
Welcome to the future: AI-generated vocal clones of superstars now available on YouTube Shorts - thanks to Google’s ‘most advanced music generation model to date’
Dubbed Lyria, and built by Google DeepMind, the tech giant says that the AI model, “excels at generating high-quality music with instrumentals and vocals, performing transformation and continuation tasks, and giving users more nuanced control of the output’s style and performance”. Google also released details of two new AI ‘experiments’ in partnership with YouTube, the first of which is dubbed Dream Track, an experiment in YouTube Shorts.
👆🏻Hot take: Google should be commended for its proactive approach to AI in this space. Of course, a cynic might argue that this is because in all cases, the platform still gets the content and therefore the ad dollars etc. Nonetheless, a smart move and a refreshing change to that of the likes of Stablility AI, below
Newton-Rex had some brisk words for his former company – and its peers – on their desire to see AI training considered fair use. “Today’s generative AI models can clearly be used to create works that compete with the copyrighted works they are trained on. So I don’t see how using copyrighted works to train generative AI models of this nature can be considered fair use,” he wrote, while offering an additional moral reservation.
👆🏻Hot take: IMO this once again lays bare the awkward attitude Big Tech has toward copyright. Quite the negative blow to Stability AI too.
Earlier this year, Spawning raised three million dollars in venture capital. The company is currently working on a handful of experiments. In October, it launched Kudurru, an open network of Web sites that aim to identify, and block, Web scrapers. Next year, Spawning plans to launch a kind of marketplace called Source+. An artist such as Bruce Springsteen could gather his data—demo tapes, vocal snippets—and license it to a company such as OpenAI, for training purposes. He could also create a model based on that data—Boss+—for other musicians to collaborate with, for a fee.
👆🏻Hot take: a terrific long read from New Yorker, with great insights into how Herndon and her business partner Mat Dryhurst are working to make AI ethically sound, thus further empowering artists. Well worth your time.
According to Spotify, “Nearly 70% of … revenue is paid back as royalties to rights holders, who then pay the artists and songwriters, based on the agreed terms.” If Spotify increased that percentage, artists would make more. But, as previously noted, Spotify is not paying artists and songwriters directly. They are paying labels, distributors, and publishers. Those entities distribute money to artists. If those entities decreased their take, artists and songwriters would also get paid more
👆🏻Hot take: a great, pragmatic look at the Spotify payment thresholds debate courtesy of . There’s a few solid reminders in here that Spotify is not the only factor in what artists earn. (Also highly recommend subscribing to Chris’s own newsletter - it’s fantastic)
By masquerading as legitimate music rightsholders, two men managed to extract over $23 million in revenue from YouTube's Content ID system. Both were arrested and sentenced to prison. In the wake of the criminal proceeding, hundreds of disadvantaged artists came forward, and the court has now ordered the scammers to pay $3.3 million in restitution.
👆🏻Hot take: good to see some crackdowns here as most of us in this space have been aware of bogus claims before now.
Stories from the Broader World of Tech:
The first, dubbed “Emu Edit,” will allow users to “precisely alter images based on text inputs.” The video demonstration for this looks similar to existing tools provided by Adobe, Google, and Canva, providing a way for users to remove or replace objects and people from photographs without any professional image editing experience.
👆🏻Hot take: the examples here are stunning, particularly the degree to which the AI can recognise specific details within the image without much effort.
It’s not clear why Clyde is suddenly shutting down. It’s possible the chatbot may return as a paid Nitro-only feature in the future, or perhaps Discord has learned enough from its testing period and decided an AI chatbot doesn’t need to be baked into its service. We’ve reached out to Discord to comment on the closure, and we’ll update you accordingly.
👆🏻Hot take: I wonder if cost might be a factor here, as bots are certainly expensive to run.
Need something else to read? Here you go:
Spoiler: They didn’t always get their wish.
👆🏻Hot take: some of these I knew, but others were certainly a surprise…
And what are the non-drinkers drinking?
👆🏻Hot take: as someone partial to a non-alc beer or two, I sometimes wonder if this is only a marginal gain. Less alcohol, maybe, but is it healthy to drink a couple of non-alc beers a night?
Thanks for reading Network Notes! Subscribe for free to receive new posts and support my work.