🔵 Death by a thousand bots
The new agentic loop that never stops releasing AI-made music
Anyone following industry news in the last 6-12 months is already aware of the degree to which AI generated music is now flooding onto DSPs. However, if it was problematic before, I would argue a crucial tipping point has just been reached, one which will escalate this from a deluge into a veritable tsunami that I can only see as a gigantic problem for DSPs and anyone attempting to use them.
Last week I became aware of this, a music MCP that allows anyone to generate songs, distribute them to DSPs and then monitor that release across platforms, all without human intervention.
(For those unfamiliar, an MCP is essentially an AI-centric programming interface in which AI agents can then interact with existing platforms in a manner not dissimilar to a human.)
In this case then, it simply means that I could leave a machine creating endless Gen AI music and then have that automatically uploaded to DSPs through a distribution platform, after which I can monitor them for monetisation. Prior to this, the upload process at least required some kind of human input.
Not any more.
If you are wondering which distribution platform is making this possible, it would appear the answer is Ditto Music, which evidently has some kind of API that ensures this is all possible.
To be clear, I do not know if Ditto are aware of this MCP and its usage, or whether such use is a breach of its T&Cs. I would imagine however that in the new age of AI, any system that has an API allowing automated distribution, including (one assumes) the creation of the account for distribution in the first place, is going to see this problem occur.
Ultimately, the news that we now have agentic AI both creating and then delivering music into DSPs should not be a shock to anyone. This was always going to happen, and it’s something many people have talked about for some time.
The bigger issue is what this means for DSPs and indeed the music landscape in general. After all, if systems are allowed to flood DSPs with AI slop, there surely has to be a tipping point in the signal-to-noise ratio wherein everything just becomes a cluttered mass of uselessness.
This long tail play from whoever is engaging with it clearly works for the person in question, but absolutely does not work for streaming platforms. Consequently, this highlights just how important it is going to become for DSPs to be able to separate a mass of low quality AI slop from either entirely human-made, or at the very least, “human-guided” music that stands a solid chance of generating a significant number of streams, acting as valuable content on the platform.
For this reason then, the issue of provenance around music - that is, whether it was created by AI or by humans - is going to become a burning issue that DSPs face.
One platform I have seen that certainly seems to be taking the right approach is Genotone. What this aims to do is provide a fingerprinting system allowing artists to declare their work as human made when uploading it into DSPs.
To be clear, this does not have anything to do with piracy; the aim is not to track music and its use. It is more a verification system allowing DSPs in particular to understand which music is certifiably human and which is not, at the point where it is being ingested into the platform.
To me, the appeal for DSPs seems obvious. I would imagine hosting costs for one are only spiralling upwards as more and more slop gets uploaded. Consequently, whatever the cost of a service like Genotone (within reason), it stands to reason that savings would be made relative to hosting alone, if not streaming bandwidth and more.
Doubtless the same types of questions ultimately present themselves regarding where the line is drawn between what is AI and what is not. I feel this is where Genotone carries value because it would clearly speak to artists who are honestly crafting their art, versus the kinds of people using the music MCP I linked to above to just churn out endless amounts of crap.
All that being said, I have seen simpler solutions mentioned previously that feel equally effective. I believe it was Martin Mills who suggested that if there was a simple low cost to keep a track on DSPs each month - let’s just say one dollar a month - then that would succeed in eliminating the lion’s share of AI slop from these platforms.
Any which way, we have now crossed the Rubicon into a space where agentic AI can flood DSPs with endless amounts of rubbish in a “monkeys and typewriters”-style attack, all presumably with a view to making money with the odd success.
Let’s hope the industry collectively gets a handle on it ASAP.
Have a great day,
D.
🎶 Listening to “An Actor Is The Only Person...Who Believes In The Words He Says” by Kirill Matveev. This is my kind of spatial, dub-infused techno, released on the perma-reliable Vuo Records label from Finland. Definitely one of those labels where you find yourself on a deep dive into the catalogue, almost all of which is absolutely superb. Check it out.
📺 (Finished!) watching “Small Prophets”. Mackenzie Crook’s latest comedy is about a man growing prophesying homunculi in a bid to find out whether his missing girlfriend is still alive. That might sound bleak, but the show itself is a warm, wonderful, easy watch, which fans of Crook’s last show Detectorists will be all too familiar with. It’s the best show I’ve watched in a long, long while. Make time for it.
🤖 Loving the Ableton Note update. For those unfamiliar, Ableton Note is the company’s mobile app that allows you to sketch tracks together with ease when on the go. The smartest aspect is the way you can sync that to the cloud and then import it into the main desktop version of Ableton Live to flesh out the song. In this latest update, though, Ableton Note introduces a load of features around audio, including slicing, stretching and more. Very smart; I’m keen to see where this will all lead.
Gleefully accepting music recommendations!
I do really enjoy getting suggestions for either existing music or forthcoming releases to check out. So, if there’s something you think I might like, do feel free to get in touch. Always keen to hear the weird and wonderful things going on out there. And, with 6000+ subscribers, I’m happy to spread word on things I’m loving too.
What I am stating here is that we must not call it music. For me It should be separate from human toiled music/songs. Generation is not producing. Also that bigger tragedy is that the Suno's/Udio's have just ripped the work and now it is generating pablum. This is pushing the expansion of art but has just made creativity industrialized. The stuff should not be called music. It is just 'sound generation'. We as a society have just lowered the bar to feed Silly - Con -Valley impersonating innovation which is word that has been bastardized!
I sense we are feeling a large force at work. The internet as we know it is breaking, as it wasn't built for this. Perhaps it's not a bad thing. Perhaps we should be far more rooted in the physical and our communities. I'm not suggesting it will disappear - I am sure there will be high trust networks - but that people's priorities might change. Why spend so much time in a place flooded with illusions and lies? How good can that be for us?