As much as illustrators are crying over AI, I can't even imagine what's going through the mind of musicians right now. AI music is on its first iteration and it's already BTFOing the current iteration of AI art. It probably doesn't help that music is fundamentally formulaic.
@RustyCrab people prefer lame music, so if you in the crappy music business, you in trouble. AI can't understand my tastes and it can't do what I do. if I had 1 fan, I might be shook, but they can't take the magic on those tapes from me.
@RustyCrab iirc logic pro x (release date 2013) came with a drum kit designer which I believe was AI-based (trained on different famous drummers to give you an authentic version of their style)
@RustyCrab eh, as with all AI things they can get it 80% right but will never close the remaining gap. Cause human intelligence is a bit more than just a statistical model.
The progress we are seeing is from like 40% -> 80%. People assume it will just continually get better. I believe there is a hard wall that current solutions will never be able to pass by their limited design.
@inertia I'm trying to formulate why I disagree with this. For absolutely everything else, you're correct.
For music I think it's different. Not because AI is just godlike at music fundamentally, but because music is so stagnant that 98% of everything just sounds like the same pop slop. The parameter space is so narrow that it has closed in on perfecting it extremely fast.
So I guess the answer is, "you need to start actually innovating if you want to keep your job"
@RustyCrab maybe the instrumentals? Depends on the song obviously. A metronom could write a rap-beat lol.
Vocals absolutely not. We cant even get normal talking right.
For everything in between... I think there is a lot that goes into a song that isn't just melody. Things that people don't consciously notice. Like the metal song you posted, lacked a certain texture or identity. It was too generic to be memorable.
@inertia
>It was too generic to be memorable.
that describes nearly all music normies listen to.
It's difficult to tell how far it will go because 40->80->[WALL] issue but my gut tells me in this particular case they'll be able to get over it. Voices are already extremely good at emoting.
Semi-schizo thoughts but I noticed that suno.ai posts do not get the attention that other types of AI art gets. I actually think this is because it's so good that it lacks the charming "weirdness" that AI normally produces. It DOES just sound like off the shelf generic music. That's exactly why it's scary.
@RustyCrab damn. Alright that moon song is pretty good. Only outro and intro suck.
Make me wonder how long a human would take to make such rather simple song. Like if it's only a few hours than AI isn't really a threat? Cause the output of the human is 100% artifact free and can be adjusted later on.
@RustyCrab@inertia I wonder when (and if) AI will be good enough to make something memorable. Good music is unpredictable, but ultimately it's still formulaic.
@Inginsub@inertia Continuing from @PurpCat post, I firmly believe that all top 40 music is just paid for through business contracts and the public is effectively instructed to like it by being forced to hear it over and over. If you did the same thing with AI music, maybe a version from next year, I think you'd get a similar result.
As for making actually good innovative music... hard to say. It may be a matter of pumping out 4,000 songs and having one guy listen to them for a good beat.
@RustyCrab@Inginsub@PurpCat yeah I've heard that too. That they coordinate radio stations and what not to play a new song over and over. Hell, thats why songs are so repetitive within. If the song is just 5 times the same things, than they effectively 5x song plays in peoples minds lol.
You're 100% correct on "popular" music. The music industry has been using algos to produce music in a given genre that is paletteable to the consumer base; the rest is marketing the artist/band, along with every radio station having essentially 90% the same playlist as every other.
There's no need to "find" the next big thing when you orchestrate the environment for it. Look at that stupid ginger from last year - he was picked up and plastered for marketing bucks and when the profit was turned, completely ghosted off social media.
@Red_Hat@Inginsub@PurpCat@inertia I'm not going to pretend to know anything about music production, but it seems to me that any sort of popular music is optimized towards minimizing any sort of "acquired taste" sounds that you get with indie bands. It needs to instantly be "easy on the ears" to appear competent, no matter how bland it is. Past that it's just the popular channels squeezing blood from the stone repeatedly. Normies don't seek anything outside of those channels so they've likely never even heard anything interesting.
@Myshkin@Inginsub@PurpCat@Red_Hat@inertia I should spend the day getting redpilled on how such an aggressively boring band got their songs played every second of every day for eternity
I thought they fit in with this convo. Lame band using a formula to make every song sound the same. Some noticed some didn’t.
I’m pretty sure Fm radio is pay to play.
@Myshkin@Inginsub@PurpCat@Red_Hat@inertia from what I understand, payola is illegal but they get around it by offering "promo packages" to radio stations who play nice, which may or may not include metric shitloads of cash.
But that was radio stations. All of that is probably completely irrelevant in the streaming era and they likely just pay4play directly.
On the other hand, as I've mentioned many times, I literally have not heard one single song made after 2010 played anywhere in public. Even in places that specifically cater to college students. That makes me think something is completely fucked on the business end of syndication.
It has a lot of good tips that just need to be modified for today, but it's a good look into how a pop song is made with lots and lots of commentary on chart toppers of the 80s with names you've probably forgotten or never heard of.
Here's an example the book mentions too, a chart topper group from the 80s known as Stock, Aitkin, and Waterman. There were similar groups in the USA as well but over in the UK they were manufacturing pop stars. Never Gonna Give You Up was not only one of their big produced and written hits, it's even mentioned in the book as an example.
There were plenty of groups doing the same shit in the USA with different stars, and also groups who would put together pop groups similar to K-Pop groups today such as NKOTB, Backstreet Boys, N-Sync, One Direction, and more. Many times you'll read about how they just wanted some dancers, they trained them to sing, and wrote all the slick chart hits for them already.
The same goes with idols in Japan and the western versions like with Taylor Swift, Madonna, Britney Spears, Lady Gaga, and the like.
@inertia@Inginsub@RustyCrab@Myshkin@Red_Hat It's even funnier when you realize that one of the writers actually made an album with these guys and Bill was the manager, so these guys weren't just LARPing as knowing shit either and analyzing it from an outsider POV.
@RustyCrab@Inginsub@Myshkin@Red_Hat@inertia also some edgy kids saw the KLF as fucking subversive m8 for the shit they did at the one award show by firing a machine gun into the crowd after doing a grindcore version of their hit single (not a joke).
A lot of their antics were either really funny/over the top, or art hoe tier shit. There was really no in between, and given what happened after this they really ran out of steam and had no idea what to do afterwards.
There was the burning a million pounds for the sake of it after deleting their entire back catalogue. A few years after "retiring" they did this weird one off reunion show where they pretended to be old people in mobility scooters to take the piss out of aging rockers clearly in the retirement home doing just one more show (we promise bro) this time for money.
Then Jimmy Cauty fell into obscurity as an art hoe (male).
@Red_Hat@Inginsub@PurpCat@inertia that's a good point, but at least normies will typically crave weird or exotic foods sometimes. I think the average person goes their whole life unaware that more than 20 songs were performed during the entirety of the 1970s
@RustyCrab@Inginsub@PurpCat@Red_Hat@inertia That's a major reason AI will be very good at pop music; pop music lives on predictability. Next time you're listening to a pop song you aren't familiar with, try to think ahead of what the next few seconds will sound like. I was pretty surprising the first time I found out I could intuit 90% of a song seconds before hearing it.
I'm actually surprised at just how much AI music skirted under the radar for so long. For instance, this video is from OpenAI's Jukebox and it came out in 2021. I guess the relatively long generation times and lack of easy to use tools contributed to AI music being forgotten in the public eye.
And there's also model poisoning where eventually the AI models will start "learning" from the output of other AI models generally decreasing the quality and creativity of the output
Add comment