Rick Rubin on AI (& now Graeme Revell, too)

[Alchemedia]

Punk, in that you don't have to know what you're doing, you just have to know what you want to express.

I'd argue the standout Punk and proto-Punk rockers instinctually knew what they were doing and how they wanted to express it.

[Opaque]

Re-creating or imitating the sound of punk-rock with AI would be ridiculous - it amounts to the averaged-out form which arose from the punk-rock spirit

Again, you're looking at it too literally. It's about ideas, not output. I don't even understand where this ridiculous idea that anything is "averaged out" by AI even comes from.

I'm not necessarily opposed to AI, but the punk-rock equivalent will be the things created with AI primarily as an instrument, not a tool - so things that are unique to it, and could only have been generated by it. I've heard already some DJ sets with AI that are crazy.

You're talking about the actual output, Rubin is talking more about the ethic, the freedom for anyone to realise their goals, regardless of skill level.

This AI is actually LLM, and it works as an incredibly advanced pattern recognition over large amounts of data, and then recreation based on that, building the syntax through probability, and all of that depending on keywords/instructions (in general, as I understand it).
That's why it's so good at doing anything that has been formalized - writing massive amounts of code for familiar stuff, using formal language for papers and structuring them according to academic standards (but actually being relatively poor in actual interpreting and reading between the lines), searching for massive amounts of data or sources and providing answers, analyzing medical documentation, rtg scans, etc. for minor deviations from the usual patterns - and that's also why, in music, the firs genres it nailed are those that are quite formalized in their execution, so country and some forms of EDM.

So what I'm saying is that it generates in relatively small variations around this imaginary perfectly averaged-out central point, guided in any direction, to any further specifics by keywords and instructions. But, one cannot be creative with guitar playing if there's no guitar, so when pushed towards anything new and unexplored it will start breaking apart, which is exactly the point when it stops being used as an advanced sample library (e.g. Splice) and is used creatively as an instrument.

The second part of your argument relates to this former usage, as a source of tracks/samples/sections, but that can already be done at the moment with e.g. Splice, although this expedites the process and makes it more flexible. So, one can already find and layer punk samples/tracks to realize their idea and create a punk song the same way Espresso was built, which is why I said that it was essentially pop at that point, and has little to do with the punk ethos.
But, for example, we also recognize that the creative use of sampling, sampling as an instrument, was not in layering of generic samples to form endless variations of established genres, but in its case in the possibility of recontextualization (which led to new genres like hiphop and jungle) and advanced editing - so in the works of artists like Daft Punk, J Dilla, DJ Shadow, Jan Jelinek, Burial, Matmos, etc.

[Michael L]

How is that (AI) any different from using a synth?

Obviously they are exactly the same: both are Artificial, neither are Intelligent.
No need to justify AI with Rick Rubin's new marketing strategy to stay relevant; make up your own, like "I prefer the company of AI bots to most Queenslanders."

the creative use of sampling, sampling as an instrument, was not in layering of generic samples to form endless variations of established genres, but in its case in the possibility of recontextualization (which led to new genres like hiphop and jungle) and advanced editing - so in the works of artists like Daft Punk, J Dilla, DJ Shadow, Jan Jelinek, Burial, Matmos, etc.

Recontextualization gets right to the point, the humanity of using cultural symbols to connect your work to others.

[jamcat]

No one works alone. Does it really matter if your bassline was laid down by a session bassist or AI? Or if your string arrangement was done by an orchestrator or AI? Historically, these are things that would be outsourced when making a record.

Likewise, does it matter if you use AI for programming grunt work instead of hiring a code monkey to do it? The purpose is the final product, not the process that got you there.

[Scotty]

It depends on how much agency you are prepared to surrender and how much of “you” you are prepared to subtract from the claim of ownership. Music creation can be purely functional. If it gets the job done, then why not? That is your central argument. That is hard to dispute, but I think there are some distinctions that should be made in your scenario.

Hiring that session bassist is in part prompt-driven, but it isn’t neatly contained within a chat. You likely heard their playing and selected them because of skill and the potential to augment the song with their input. It likely wasn’t a blind handover. Their performance wasn’t perfect. There were some happy accidents that you maybe exploited in the editing. If you were asked about that bassline, you probably could describe why it worked, why you chose that particular take, and what that bass player’s contribution meant to the song. In a sense, you could defend the work, or at least speak to it. Maybe that doesn’t matter. But if we don’t acknowledge that the process itself has value, not just the outcome, we risk losing the ability to meaningfully relate to the work at all. It becomes harder to explain why something worked, harder to stand behind choices, and easier to treat creation as a series of simple outputs.

I was at a celebration of life recently. A person spoke eloquently about the deceased. They overcame their fear of public speaking and remembered their friend by reading a script that ChatGPT had made for them. Privately, they also volunteered that they were under some time pressure, so they needed the help. It was an achievement, no doubt. A person who likely wouldn’t have spoken found a voice. It wasn’t her voice and it most likely wasn't her words.

Despite the obvious benefits, I think that remains a distinction with a difference.

No one works alone. Does it really matter if your bassline was laid down by a session bassist or AI? Or if your string arrangement was done by an orchestrator or AI? Historically, these are things that would be outsourced when making a record.

Likewise, does it matter if you use AI for programming grunt work instead of hiring a code monkey to do it? The purpose is the final product, not the process that got you there.

[machinesworking]

I think it really comes down to the difference between a specific person with their own hyper critical tastes who comes up with some slightly unique style based on those tastes. AI is not at this point capable of discernment, it likely never will be. So you put in post punk bassline and you will get what the programmers tagged as post punk basslines they used to machine learn how to play that genre, it will be a distillation of all of the style of that instrument in that genre without any discernment based on ancillary genres that person loves.

Almost every great musician I know does not stay in their genre, they love a couple completely unrelated ones. Those subtle influences from that wider taste make them unique players. I think AI will do a good job at coming up with boring basslines and drum patterns and backing tracks in general, but if the individual doesn't input some of their own weirdnesses into the main parts of the song, it's just going to dilute music into bland remakes. Sort of like how every genre ever starts becoming archetypical and loses some initial shock of the new, but times 100.

[BONES]

I'd argue the standout Punk and proto-Punk rockers instinctually knew what they were doing and how they wanted to express it.

If you bothered to read in context, which I included in the quotes, the meaning would be clear. That said, Glen Matlock was the only member of the Sex Pistols who knew anything about what he was doing. Yes, Steve Jones apparently had some natural talent with guitar but no formal musical training of any kind. But if you weren't around at the time, you'll have no idea what it was like, of the thousands of bands that popped up out of nowhere, emboldened by the Punk ethic of just getting up and doing something, of expressing yourself.

This AI is actually LLM

Which AI? we're not talking about any specific AI.

... it works as an incredibly advanced pattern recognition over large amounts of data, and then recreation based on that, building the syntax through probability, and all of that depending on keywords/instructions (in general, as I understand it)...
... and that's also why, in music, the firs genres it nailed are those that are quite formalized in their execution, so country and some forms of EDM.

You don't think that might have had more to do with popularity of genres? Popular genres will have more training material available and will be far more likely to be asked for by users. Sora, for example, had no clue what EBM or Industrial music even was the first time I tried to use it. It had probably not been trained on any of it to start with. But Tunee seems to be pretty good at it today.

So what I'm saying is that it generates in relatively small variations around this imaginary perfectly averaged-out central point, guided in any direction, to any further specifics by keywords and instructions.

You mean it uses the average as a starting point so it doesn't have to search further in one direction than another. Surely that's just common sense? And it's just a different way of saying what I've already said. The key point is that you use your prompting skills to move it way from the average to the area where the music you want to make lives.

But, one cannot be creative with guitar playing if there's no guitar, so when pushed towards anything new and unexplored

That's a very specific situation, there are others where it will explore in directions no human would think to go. The example I've used in previous discussions is that Tunee has combined the European music of EBM with the American vocal style of Electro-Industrial to come up with something that feels quite new and fresh. It's a sound I've not encountered in the 30+ years I've been working in the genre. So if it can't do things that can be done in other ways, who cares? But if it can do things a human might never think to do, then it's opening up new and unexplored possibilities that might otherwise never see the light of day.

it will start breaking apart, which is exactly the point when it stops being used as an advanced sample library (e.g. Splice) and is used creatively as an instrument.

Or when it just becomes useless garbage, like most music that is trying to be unique in a world where everything worth doing has already been done a thousand times. Ai seems to do what we've always tried to do, it finds the spaces between where others have gone before and makes things that might not be breaking new ground but are definitely unique in their own way.

So, one can already find and layer punk samples/tracks to realize their idea and create a punk song

Nobody is talking about creating Punk songs, Rubin was simply likening the rise of Ai to the rise of Punk as a movement, not as a genre, how it can empower a whole new group of people who might otherwise feel they have no means of expressing themselves.

it was essentially pop at that point, and has little to do with the punk ethos.

Those are two largely unrelated things. Buzzcocks are remembered as one of the most successful bands of the Punk era but much of their music falls into the Pop genre. Punk empowered them to get up and express themselves, it didn't determine how they should sound or what their music should say. But, just as it's been for me, if Punk hadn't come along and showed them, they'd never have believed they could get up on stage and play music.

No need to justify AI

Nobody needs to justify what they do, the work will stand on its own or it won't. We've already had ample evidence that AI generated music works, that people will listen to it and enjoy it.

the creative use of sampling, sampling as an instrument, was not in layering of generic samples to form endless variations of established genres

Neither is using AI. It just doesn't work like that. It works a lot like we do - it takes "inspiration" from other sources to produce something of its own. It's not cutting and pasting, it's making music from scratch.

[BONES]

It depends on how much agency you are prepared to surrender and how much of “you” you are prepared to subtract from the claim of ownership.

Quite the opposite, it's about how much you are prepared to put in to ensure that what comes out is exactly what you were trying to achieve. It's the same with any process. Do you use a guitar tuner or tune by ear? Do you use an arpeggiator or play every not by hand? Do you use a chord tool or work that stuff out for yourself? Do you build your own instruments or buy the result of someone else's work and claim what you make with it as your own? Do you keep doing takes until you get it right or do you edit your playing in the piano roll and pass it off as your own playing?

Hiring that session bassist is in part prompt-driven, but it isn’t neatly contained within a chat. You likely heard their playing and selected them because of skill and the potential to augment the song with their input.

You also get to choose an AI before you get into any chat environment. And I don't see any difference between a written chat with an AI and telling a session muso what you want him or her to do.

It likely wasn’t a blind handover.

Nor is it with AI.

Their performance wasn’t perfect. There were some happy accidents that you maybe exploited in the editing.

Which you'll get x100 when working with AI.

If you were asked about that bassline, you probably could describe why it worked, why you chose that particular take, and what that bass player’s contribution meant to the song. In a sense, you could defend the work, or at least speak to it. Maybe that doesn’t matter. But if we don’t acknowledge that the process itself has value, not just the outcome, we risk losing the ability to meaningfully relate to the work at all.

All of which is at least as true when working with AI, much of it moreso. It's just a different process. If you've spent 50 hours trying to craft prompts to get what you're after, you'll have thought far more deeply about what it is you are trying to do than you ever would working in more traditional ways.

It becomes harder to explain why something worked, harder to stand behind choices, and easier to treat creation as a series of simple outputs.

no f**king way! It's the exact opposite because to get what you want, you have to be able to articulate it to the AI. OTOH, most of the time you are making up justifications after the fact when you are working the old fashioned way.
quote]A person who likely wouldn’t have spoken found a voice. It wasn’t her voice and it most likely wasn't her words.[/quote]
You mean like when someone reads a bible verse or quotes from a poem at a funeral or wedding? Are those moments any less potent? No, often they add extra potency to proceedings.

I think it really comes down to the difference between a specific person with their own hyper critical tastes who comes up with some slightly unique style based on those tastes. AI is not at this point capable of discernment, it likely never will be. So you put in post punk bassline and you will get what the programmers tagged as post punk basslines they used to machine learn how to play that genre, it will be a distillation of all of the style of that instrument in that genre without any discernment based on ancillary genres that person loves.

If that was true, wouldn't it always spit out the same thing every time it got that prompt? It won't. Try it 10 times and you'll get 20 very different results.

Almost every great musician I know does not stay in their genre, they love a couple completely unrelated ones. Those subtle influences from that wider taste make them unique players. I think AI will do a good job at coming up with boring basslines and drum patterns and backing tracks in general, but if the individual doesn't input some of their own weirdnesses into the main parts of the song, it's just going to dilute music into bland remakes.

That's up to you. If you tell the AI to give you a PP bassline (although f**k knows what that even means) but in the context of an Industrial Rock song, it can do that. It's not tied to genres, it does what you tell it to do. And Tunee comes up with way more interesting drums than I do. That's one of the good things about working with AI. Drums for me are always an afterthought, the last thing I do and something I put comparatively little effort into but Tunee doesn't know that so it comes up with some great percussion.

[Michael L]

No need to justify AI

Nobody needs to justify what they do, the work will stand on its own or it won't.

AI work never stands on its own.
It stands on the average of everyone else's work.
It's this era's Muzak, and yes 50,000 new AI-generated tracks are now uploaded daily, up from 10,000 a day just one year ago, so you are totally on-trend, per usual!

[imrae]

He likens what he calls "vibe coding" to Punk Rock - "So in the past, for music, you had to go to the conservatory and study for years and years. Then someday, you could play in a symphony. And then, when punk rock came along, you could maybe learn three chords in a day — and there were all these bands."

This is a ridiculous statement that disregards the whole of folk music. People were learning a handful of chords and expressing their feelings for hundreds (likely thousands?) of years before punk rock came along.

[revvy]

Old dogs, new tricks. Who knew?

[Bunny_boy]

Is this punk rock in the same way that Brew Dog goes on about being punk?

[telecharge]

This is a ridiculous statement that disregards the whole of folk music. People were learning a handful of chords and expressing their feelings for hundreds (likely thousands?) of years before punk rock came along.

"Ridiculous" seems a bit harsh, but I think I get where you are coming from.

I understood it to be a 'counter cultural' analogy about not needing to go to school and 'put in the hours' before getting to work.

Is this punk rock in the same way that Brew Dog goes on about being punk?

Never heard of them, but I wish people cared about the blues the same way they do about punk.

[telecharge]

In case you missed it, Instagram boss says ‘more practical to fingerprint real media than fake media’ - https://www.engadget.com/general/the-mo ... 44371.html

[Bunny_boy]

Is this punk rock in the same way that Brew Dog goes on about being punk?

Never heard of them, but I wish people cared about the blues the same way they do about punk.

I don't think the blues is seen as being rebellious to readers of GQ and Vogue in the same way that punk is.