Be advised: this thread is not for serious in-depth discussion of weighty topics (we have a link for that), this thread is not for anything Culture War related. This thread is for Fun. You got jokes? Share 'em. You got silly questions? Ask 'em.
- 160
- 4
Jump in the discussion.
No email address required.
Notes -
Hey guys, remember a month and a half ago I pointed out that AI-Generated Music had fully crossed the uncanny valley?
I specifically claimed:
GUESS WHAT.
https://www.whiskeyriff.com/2025/11/08/an-ai-generated-country-song-is-topping-a-billboard-chart-and-that-should-infuriate-us-all/
I do think this either proves that the average country music fan has little taste, or AI music is as good or better than the average country musician.
Damning with faint praise, perhaps, but this absolutely still feels like we've officially entered a new state of play for the music industry.
Just before AI music became a thing, Ted Gioia talked about a Spotify fake artist problem he discovered. Bascially, he noticed that playlists with titles like "Jazz for a Rainy Afternoon" wouldn't include any artists that he recognized (and as a jazz critic he would recognize more than the average bear), and further investigation revealed that the "albums" the songs were from would only have one or two songs. Looking into this even further, he discovered that he couldn't find much information at all about these artists, except addresses in the Stockholm area. The conclusion he came to was that since some music styles—jazz, chillout, orchestral, etc.—are driven more by algorithms than individual artists (by virtue of people telling Alexa "Play relaxing music" or whatever), it was cheaper for Spotify to hire studio musicians to record generic slop so they wouldn't have to pay royalties to real musicians.
In the 90s, I was at a discount store with my dad during Christmas season and he bought a CD titled "Jazz for Christmas Eve" for a dollar or something. It didn't have the name of any purported artist, just song titles. The music was entirely MIDI. A few years later the mother of a family friend was going into the home, and we were helping to clean out her house. I took the records, mostly junk, but there was one that stood out. It was called "The Hits of Nat King Cole" or something similar and had a picture of Mr. Cole on the cover. Towards the bottom, in relatively small print, it said "Performed by Bob Gigliotti" or whoever. The liner notes weren't extensive but mostly talked about Nat King Cole. The only mention of the gentleman who was actually performing on the album was a brief paragraph that said that he was, in fact, a singer, and that he does a good job with the material. When I played the record, I was hit with some guy doing an uncanny Nat King Cole impression.
The point I am trying to make is that cheap, mass-produced slop has existed in the music industry for as long as production costs were cheap enough to justify it. An enterprising music historian could probably do a book-length treatment of the subject, but in the end this has only been a minor footnote in the history of music. And even in the limited instances where it has historically gotten a foothold, tides shifted away from it. Consider Muzak. I hesitate to call it slop because, up until the 1980s, it was produced with a degree of professionalism and creativity that belied its status. But this was more for the pleasure of the people making it than anything else; it was always intended to be nothing more than musical wallpaper for stores, offices, and other public places, with orchestral arrangements of popular hits almost algorithmically selected to ensure the proper pacing. In the 1960s it was ubiquitous, but these days the only national chain I can think of that still plays this kind of music is Hobby Lobby. Retail started shifting to name artists in the 1980s, starting with inoffensive "soft rock" but more recently including practically anything that's been popular since the 1960s.
The AI doomers have tried to make the argument that because this music can be generated so quickly and so inexpensively it's trivial to just completely flood the market, and cash-strapped record companies would love it if they could generate product without having to pay the artists, producers, etc. While this may seem like a compelling argument the music industry could have always done this, but they haven't even attempted it in 100 years of existence. Making music is obviously a skill, and making music that people want to listen to (and pay for) is an even greater skill, but it's not a particularly unique skill. Any city is going to have hundreds of musicians who write their own material, practice in their spare time, play in bars in the weekends, and are good enough that most of the people in attendance enjoy the performance. If the record companies wanted to, they could have always signed as many of these musicians as they could, pay for a recording session, pay the musicians a low flat fee, and completely spam the market for little cost. If they get a hot or two out of the deal, great. If not, they're only out ten grand.
In reality, major labels are highly selective about who they sign, and those they do sign usually get significant financial backing. A local band recording at a local studio can get an album out the door for about 5 grand if they're well-rehearsed. A major label will spend, on average, $250,000 to $300,000 to record the album. The label will also pay for promotion, which can run into the millions if touring is involved. And they would always prefer to spend money on a proven star rather than a nobody. In other words, the model they operate on is the exact opposite of the one where AI takes over.
And it gets even worse. In an alternate universe where record labels operated by signing cheap labor and spamming the market, that at least allows for the possibility of being able to capitalize on the hits. AI doesn't even allow that, since there's no guarantee that you'll get output that's plausibly by the same fake band. Even big stars like The Beatles, Frank Sinatra, The Supremes, and Taylor Swift have had occasional flops; there's no guarantee that because an artist is popular that any individual release will be successful. But at least when you got The Supremes in the studio you were guaranteed a Supremes record. With AI you just have to keep generating and hope that you eventually get a Supremes record, and even that doesn't guarantee you anything.
As much as AI doomers talk about how it's going to take things over, it's not. It's going to replace slop, but slop has always existed. The business model doesn't really allow for the kind of dystopian future they're predicting.
I don't think I fully understand what you're getting at here. I mean I get the idea that record companies always had the option of hiring studio musicians to churn out whatever, but the effective cost to produce and distribute that kind of thing was never zero, the way it is now.
Get some random band together, put them in a studio, let them work for weeks, and if it sucks you're out "only" low five figures? What kind of Stone Age shit is that? Your competition is a guy with a laptop, his product is basically indistinguishably as good as yours at this point, and oh yeah he has no expenses, can never go out of business, and there are endless thousands of him.
If you're a record executive who has been given a 1 million dollar budget to develop an artist that the label expects to have a hit, what do you think is the better strategy?
Scout someone whom you think has potential and spend the 1 million signing, recording, and promoting them.
Sign 100 bands more or less at random and spend $10,000 signing and recording them, and just release the music and hope it promotes itself. You're going to pay them up front for the rights to the recordings so they're just making an album for hire and aren't under contract, and you won't even bother to keep their phone numbers. What they record now is it.
The artist you spent the whole million on might make the money back and might not, but the chances of that happening are much better than assuming that one of the hundred albums you just threw into the market with no promotion is going to be a big enough hit to recoup the costs of all the others is practically zero. The calculus doesn't get any better if you can record 200 artists for $5,000 apiece, or 400 for $2,500 apiece. At that level of investment it's akin to a lottery, and lotteries don't become better investments just because the price of tickets is cheaper. It's actually worse than a lottery because at least the lottery has calculable odds and a guaranteed winning number.
Acquiring the rights to potentially popular music and gambling on it in the form of promotional dollars will remain a viable business model, but that's really all we're talking about here. The "record company" of X years from now may well be one guy with a squad of AI agents trying to get his hookiest AI songs into the right TikTok cat videos.
More options
Context Copy link
More options
Context Copy link
Yes it was, because people were willing to do it for free and studio space has also been available for free for decades.
Distribution became free at least 15 years ago and well before AI.
The thing that has been in limited supply is attention and interest, and that been true for like 50 years at least.
More options
Context Copy link
Maybe the point is that in making the next Taylor Swift probably 0.1% spend is on music (regardless of whether AI or human made) and 99.9% is on marketing, both standard and native.
Even for a mega artist like Taylor Swift, that's not entirely true. Marketing is certainly part of it, but there's a lot more that goes into it. Even if you cynically assume that pop stars are all marketing and no substance, labels pay a lot of attention to what gets released. The reason that people like 2Pac seem to have immortality is because they all record a lot more than the record company is willing to release, especially with pop musicians who use outside songwriters. If it were simply a matter of spamming the market with material then they either wouldn't record as much (to save money), or release everything they did record (to maximize revenue). The reason they don't do this is because they need to maintain a certain quality standard and avoid saturating the market. In the 50s and 60s artists were required to put out several ten song albums per year. By the end of the sixties, release schedules slackened, and by the 80s an album a year pace was considered pretty good. Now they can go years between releases, and this isn't due to lack of material in most cases; those 50s releases included a lot of filler.
So yes, they are paying attention to the music, and AI doesn't allow one to pay any attention to the music, especially when it's made by people with no musical experience. It's just spamming in hope they can make more money than they spend, with little control over the content. And marketing includes a lot more than what one typically thinks of as marketing. It includes touring, arranging press interviews, making sure critics review the album, making public appearances, having ins with radio stations, and all of that.
More options
Context Copy link
I don't claim to know how it's all going to shake out, but Joe Blow on his laptop doesn't really need to create the next Taylor Swift, does he? He just needs to capture a non-zero amount of the Music Dollars that exist in exchange for his investment of literally nothing. Once that's possible, and it seems like we're there already and only getting better, the fact that an endless supply of Joe Blows exist pretty much guarantees serious disruption in the long term.
Or maybe not, who knows, right? But when I see someone post something that sounds a lot like "sure this technology makes (thing) for free but it doesn't really suit how the (thing) industry works" my gut says that's it's the industry that's going to have to cope.
Some people capturing a non-zero amount of the market doesn't equate to upending the market to the point that it has to cope. Slop has always accounted for a non-zero amount of the market.
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link