Posted
Who would have thought that A.I. generated music would become a tool for organized crime? This story totally baffled me.
Fraudsters generated incredible numbers of A.I. "songs" and uploaded them to platforms like Spotify. Then they used streaming farms (massive bot nets that fake real users) to stream their channels 24/7. This made them eligible for millions in royalties.
When the CEO of a streaming platform tried to mitigate the fraud, he received death threats! I mean, Cognitone also received threats of physical violence from fraudsters, but if there are millions at stake, the threat is much more credible and scaring.
As for the music, I think it can be quite nice for a few minutes. After listening a bit longer though, the generic formula becomes obvious. Maybe it's just me, but the novelty wears off quickly and then it starts to feel like a brainwash. The CIA might be interested in using A.I. loops for torture.
A total of 14.4 million songs have already been generated by A.I. and uploaded to streaming services. That's about 14% of of the world's total recorded music of all time! As you think of it as a musician, it can be depressing.
Bots don't care about quality. And their masters will cut off your thumb, if you don't pay them their share.
Mon, 2023-06-12 - 22:26 Permalink
Streaming platforms will have to find a way to deal with this, or their business model is toast. No one wants to wade through a mountain of crap to hear something decent. It was bad enough before the advent of AI.
Humans excel at finding ways to sh*t in the pool.
EDIT: And of course, legit artists will see their deserved royalties decrease.
Tue, 2023-06-13 - 08:18 Permalink
Generative A.I. will inevitably get a bad rap because of this. Now it's a buzzword. At some point it might become a liability that software developers would rather want to hide.
That's a pity. A.I. is great, provided there's a human involved in the process who uses it as tool and is still responsible for the many creative decisions that are necessary to make a piece of music personal and unique.
It's impossible to make these decisions if you "never made music before" or "don't know anything about music" (as these A.I. services claim to grow their audience).
The irony is, the more people are using a generator, the more it dilutes the value of its output, eventually converging to zero. This trend will eat itself. I can't help but feel this is just a pump and dump scheme.
So it's only consequential that criminals leverage it to the max, while it lasts.
Tue, 2023-06-13 - 16:10 Permalink
My music listening experience certainly won't change with AI. Because already in the past, hundreds of thousands of tracks were uploaded to the platforms every week. Nevertheless, only once every few months I did stumble across a track that I really liked. This will probably not change if in future 100 million tracks are uploaded per week instead of hundreds of thousands. At most, it will be even more difficult to find something decent. And the radio stations will continue to play Bon Jovi all day long.
Tue, 2023-06-13 - 16:33 Permalink
"A.I. is great, provided there's a human involved in the process who uses it as tool and is still responsible for the many creative decisions that are necessary to make a piece of music personal and unique."
This is exactly what I like so much about Synfire. It presents me with options I never would have thought of, but many of them are "almost, but not quite there." So I have to work on them from that point...add/delete/move (vertically or horizontally) symbols, try different chords or synth patches, duplicate a phrase and have a different instrument play it at half tempo, sync the groove of one instrument to another by using the Rhythm parameter, vary phrases via the Transpose and/or Pause parameters, make changes on a macro level via containers, and a thousand other methods that I'm only now beginning to understand. Furthermore, Synfire makes it relatively easy to do all of these things once one gets past the initial steep learning curve.
From all this I derive a great sense of aesthetic satisfaction. This is far removed from simply pressing a button and sitting back while a finished piece of music pops out of my computer. Along the way I use my decades of experience in various aspects of music production and performance, and I learn new things in the process. I don't consider myself to be quite on a pro level yet, but Synfire gets me a lot closer to it.
Tue, 2023-06-20 - 20:51 Permalink
> Streaming platforms will have to find a way to deal with this, or their business model is toast.
The story points to the ‘pro-rata’ streaming model of streaming services like Spotify as the core problem, as it incentivizes misuses like this.
The article suggests as one possible solution to instead use a ‘user-centric’ model, where moneys from the streaming subscription fee of, say, user A is paid out only to the artists user A actually listened to (instead of distributing moneys from a single central pot).
Other models are suggested as well, e.g., Universal Music Group promotes what it calls an 'artist-centric' model. Also, "Universal may call for AI-made music that falls below minimal-quality identifiers to be blocked from being uploaded to digital services." The question still remains how to define/filter "below minimal-quality identifiers", because of course the music generators can adapt to the chosen definition.