The number of songs in the world doubled yesterday. Did you even notice?
An artificial intelligence company in Delaware boasted, in a press release, that it had created 100 million new songs. That’s roughly equivalent to the entire catalog of music available on Spotify.
It took thousands of years of human creativity to make the first 100 million songs. But an AI bot matched that effort in a flash.
And the company that delivered this feat is led by a someone who studied Jazz Bass in college… A bass player. This is odd, I always liked bass players. They seem reasonable, grounded, solid. They don’t seem like somebody proud of letting machines make 100 million songs.
Read the linked article and let me know what you think.
>A bass player. This is odd, I always liked bass players. They seem reasonable, grounded, solid. They don’t seem like somebody proud of letting machines make 100 million songs.
Most are.
Consider: ECM Producer/Founder Manfred Eicher is/was a bass player, The conductor Zubin Mehta is/was a bass player, and one of the most innovative audio engineers in the 20th c (along side Rupert Neve) George Massenburg is/was a bass player.
So many are grounded, and solid … but some… in the same way that Tom Riddle became Lord Voldemort (Harry Potter reference) some bass players do end up going to the “dark” side. There’s no explanation apart from, “They could, so they did” … as do many in the arc of history.
… or … as in the case of Voldemort perhaps the person referenced in this article is just a “self-hating bully” … I dunno … I didn’t know of this person … nor am I particularly interested in what they have done technically. Seems to me an engineering parlor trick. This is really nothing more than a permutation exercise.
From my personal experience (anecdotal evidence follows): it seems that a great many strides in AI have experienced a … what to call it … regression to the mean … in subsequent release increments since the “big splash” back in late November 2022. And this is to be expected.
I read the article, and agree with most of the concerns. However, it should be made clear that the company which generated those 100 million tracks did not do it merely as a frivolous exercise. They are an on demand soundtrack provider for podcasters, bloggets, live streamers, ad producers, etc. and those tracks were produced by those clients. The company pays sound designers for the rights to their sample libraries and loops, which are then used in the deep learning generated music compositions. While they do not sound anything near genius, I think that is not the intention.
Now, over on Spotify, if half the content is AI generated crap that listeners prefer to the already existing tracks, I say let it be, for the problem is the audience has become conditioned to the consequences of preceding technologies including digital reproduction, samplers, beat quantization, auto-tune, mp3, and streaming services. The bar is set low and the floodgates are already open.
The good news is we can still escape the masses, go out into the woods and take in the most awesome creations. And hopefully soon, backstage!
>They are an on demand soundtrack provider for podcasters, bloggets, live streamers, ad producers, etc. and those tracks were produced by those clients. The company pays sound designers for the rights to their sample libraries and loops, which are then used in the deep learning generated music compositions.
Which begs the question – why does one need to have a cadre of 100 million songs for sale/lease/use/etc?
Why 100 million? Think about the population size of podcasters, bloggers, et. al. What’s the size of the user base? 100000? 250000 ? 500000? I think the sheer number of “songs” produced actually shows that the release is in fact a technological parlor trick. There is a very definite “we did, because we can” element to this.
They are not saying they amassed a catalog of 100 million pieces of music overnight. That figure is the to date number of parametrically, on-demand, customer generated tracks (56M downloaded, 44M streamed), over the last few years.
Why 100 million? Because people tend to make special significance of 10^nth events!