AI disqualifies anyone as a musician! It's like playback.

[zerocrossing]

I love how eager BONES is to replace his own lame music with computationally generated lame music. There's a beautiful symmetry to it.

(and yes, AI is training on it's own data, and while not continuously, each model becomes more and more a product of its own slop)

https://www.nature.com/articles/s41586- ... enerations.

[tq]

AI training on its own output is an issue, you're absolutely right, as this can introduce model deficiencies in each iteration. There are remedies to this, though, and the problem is not an AI or LLM issue per se, but one based on how it is applied by some tech companies. As I argued above: For music, training a model on existing audio stems could be an interesting idea, and would improve model quality, especially for single track generation. I don't think you just need more data for model improvements, but 'pure' data. You can develop all kinds of helpful tools from this, that could have an application in studio contexts, and would go far beyond the 'one-liner Suno jockey thing' that seems to be the arch enemy in this thread.

[ksandvik]

I'm totally for more AI centric mastering, I could do it but it's one of the least enjoyable parts in music making. Or, just set the tone and make sure that it fits those criteria. Suspect Isotope Tonal Analyzer + Izotope is heading into that direction. Especially as in their current Ozone you could go in and edit their settings after their 'intelligent mode' has added the settings.

As for music making, that's another story, I have zero interest in using others' hard work as LLM patterns and spin variations out for that. You don't even need a human for that, just an AI agent spewing out thousand songs an hour.

[zerocrossing]

I'm totally for more AI centric mastering, I could do it but it's one of the least enjoyable parts in music making. Or, just set the tone and make sure that it fits those criteria. Suspect Isotope Tonal Analyzer + Izotope is heading into that direction. Especially as in their current Ozone you could go in and edit their settings after their 'intelligent mode' has added the settings.

As for music making, that's another story, I have zero interest in using others' hard work as LLM patterns and spin variations out for that. You don't even need a human for that, just an AI agent spewing out thousand songs an hour.

iZotope has had this for a while. I find it’s not perfect, but it’s a good way to get you started. I’ve never accepted it’s settings without tweaking them, but it beats starting from scratch.

[BONES]

I do the same but I often wonder if I'm not just doing that to make me feel more like it's me doing it, not the Master Assistant, rather than me actually making it better.

There are AI assistants to help people, which means there’s still humans massively involved. I’m talking about the Suno/Udio grinders who only care about $$$$

I am also talking about Suno, Udio and other AI tools designed to generate full, finished songs. They can be excellent compositional tools, extremely useful in developing existing ideas or inspiring new ones. The songs we have created with Tunee have broadened our range considerably, challenging us to push boundaries way more than we'd ever have done on our own. Kind of in the way a producer like Flood tries to push artists out of their comfort zones and try new things.

We are far more excited about this release than we were about any of our previous albums. In fact, it's probably the best thing that's happened to our music since we moved from hardware and started working ITB. Only an idiot would cut themselves off from the possibilities it offers just because someone else might use it to make a quick buck. It is so much more than that.

That is true and i do understand their worries and frustration. We all need money to eat and have a good life in this modern world.

For every job that disappears, new jobs are created. I've had three totally different careers in my life, that's just how the world works today.

each model becomes more and more a product of its own slop)

Again, that's not how it works. It doesn't average stuff out from it's entire store of information, it picks and chooses things based on your prompts. The more it trains on, the broader its palette gets.

https://www.nature.com/articles/s41586- ... enerations.

You can't base your opinions on wishful thinking. There is one little word in the article that makes it clearly a work of speculation. I have bolded and underlined it for you in the quote below -

"although current LLMs2,4,5,6, including GPT-3, were trained on predominantly human-generated text, this may change."

The corollary to "this may change" is "this may not change". Everyone will choose a side, we'll see who is correct in the long run.

[starflakeprj]

Let's make a comparison in the world of sports.

In tennis, improvements in rackets made from graphite and carbon fiber have enabled players to hit with more power, control, and spin. Enhanced strings also add durability and performance, allowing players to be faster and more precise.
- This didn't make tennis players unemployed. Why? They embraced the new equipment.

In golf, advancements in club design, particularly with titanium and graphite materials, have allowed players to hit longer and more accurate shots. The development of high-tech golf balls also contributes to increased distance and control. Players have greatly benefited from these innovations, pushing the boundaries of performance.
- Did this make golfers unemployed? No. Why? They embraced the new equipment.

In table tennis, improvements in racket materials such as carbon fiber and high-quality rubbers have enhanced control, speed, and spin. These advancements allow players to execute faster and more precise shots. Players have taken advantage of these innovations to reach new levels of performance.
- Millions of chinese players had to quit because of this? I hardy think so. Why? They embraced the new equipment.

I can continue, but to conclude:

Just like athletes embrace new equipment to enhance their performance, producers and songwriters can use AI as a tool to create music. AI doesn’t replace the creative process, but opens up new possibilities, just as improved equipment helps athletes reach new heights.

[pekbro]

Interestingly, in Golf, they are seriously considering designing equipment with reducing the performance of high level athletes in mind. They are hitting the ball too far. heh

[starflakeprj]

Interestingly, in Golf, they are seriously considering designing equipment with reducing the performance of high level athletes in mind. They are hitting the ball too far. heh

Really? I had no idea, but maybe golf would be boring to both play and watch if all holes would be par 2 holes

[Scotty]

In golf the player doesn't get to upload his swing into a GPU farm and have it produce the shot in full or in part. There is that.

Interestingly, in Golf, they are seriously considering designing equipment with reducing the performance of high level athletes in mind. They are hitting the ball too far. heh

[pekbro]

Interestingly, in Golf, they are seriously considering designing equipment with reducing the performance of high level athletes in mind. They are hitting the ball too far. heh

Really? I had no idea, but maybe golf would be boring to both play and watch if all holes would be par 2 holes

That's the story pretty much, as I've been told as someone who works in the golf industry.

[zerocrossing]

Let's make a comparison in the world of sports.

In tennis, improvements in rackets made from graphite and carbon fiber have enabled players to hit with more power, control, and spin. Enhanced strings also add durability and performance, allowing players to be faster and more precise.
- This didn't make tennis players unemployed. Why? They embraced the new equipment.

In golf, advancements in club design, particularly with titanium and graphite materials, have allowed players to hit longer and more accurate shots. The development of high-tech golf balls also contributes to increased distance and control. Players have greatly benefited from these innovations, pushing the boundaries of performance.
- Did this make golfers unemployed? No. Why? They embraced the new equipment.

In table tennis, improvements in racket materials such as carbon fiber and high-quality rubbers have enhanced control, speed, and spin. These advancements allow players to execute faster and more precise shots. Players have taken advantage of these innovations to reach new levels of performance.
- Millions of chinese players had to quit because of this? I hardy think so. Why? They embraced the new equipment.

I can continue, but to conclude:

Just like athletes embrace new equipment to enhance their performance, producers and songwriters can use AI as a tool to create music. AI doesn’t replace the creative process, but opens up new possibilities, just as improved equipment helps athletes reach new heights.

This is simplistic, and like so many other arguments misses the point. Let's use Photoshop as an example. if I use its AI capabilities to remove a telephone pole from an image of a corn field that I took, that's a tool. I could do that using the rubber stamp tool, like I used to, but it's an often long and tedious process. I spent years going it, and while I'm very good at it, it's not fun or creative. All hail to our computer overlords and the new process.

But now I can type, "replace telephone poll with giant banana." I still have some control, though maybe I have to accept what Adobe's AI decides what a banana is like. I don't control where the bruises are, or if it's got a bit of green left in it. I could try to get it close with prompts, but in my experience, it's never good.

But with something like Sonus, we can say, "make me an image of a corn field with a giant banana in it." Now you have very little control, and often the more specific the prompts become, the worse things get. I didn't take the photo, so I couldn't control the depth of field, or how tall or developed the corn was because of the time of year I took it. I can't control the time of day with any precision.

I guess that's faster, but is it better? Will I see that cornfield in someone else's image? I've already seen Midjourney image from other people that were nearly identical to what I got. Multiple times, actually. Who owns it? The first person who asked for it? Midjourney? Do I feel any connection to it, or is it just another bit of content to be shoved out there and abandoned?

Now apply that to song writing. When I write a song, it pops out of me in some raw form, and then I hone it over time, often by performing it with a band. Those people bring themselves into it, and to some degree enhance and shape the music in a way that I couldn't really predict. Then I go into a studio and start recording and producing it. I'm making thousands of choices. Each one informed by years of my own experiences and my emotional connection to my experiences, all which have formed me. I'll use some technology to make parts of the job that I can't do possible, but it's all under my very direct control. It's not a fast process, but in the end, I have a deep feeling of accomplishment. That feeling fuels my next project, and the next after that, even if no one listens to my music, which trust me, they do not.

So since no one listens, why not just ask Sunos to make a song for me, based on my instructions? Because that's not making music, and I am a musician. If that's what you want to do, no one is stopping you, but if you think you're a musician, you're delusional.

[Tiles]

Your argument draws a distinction between “using tools” and “not making music” based on control and authorship, but that distinction doesn’t hold up consistently. And again the distinction between the AI tools is missing. It is not all Suno.

First, the line between tool and generator is arbitrary. Modern tools across creative fields have always abstracted effort. A compressor, a preset, a sample library or pitch correction all make decisions for the user. AI systems simply operate at a higher level of abstraction, but the principle is the same: the user defines direction, evaluates results and makes selections.

Second, control is not a reliable measure of musicianship. Many established practices intentionally reduce direct control:

sampling and collage in hip hop
generative and algorithmic composition
working with session musicians who interpret material

In all these cases, the creator does not control every detail, yet authorship is not questioned.

Third, your argument is based on process rather than outcome. You describe making thousands of decisions as the defining feature of being a musician. But selection, iteration and curation are also decisions. A producer choosing sounds, takes or arrangements is not fundamentally different from someone guiding and refining AI output.

Fourth, the claim that AI results lack individuality is not specific to AI. Similarity exists in all tool-driven production environments, from genre conventions to preset-heavy workflows. That has never been used as a criterion to deny musicianship.

Finally, the conclusion that users of tools like Suno AI music generator are “not musicians” is a definitional claim, not a logical consequence of your argument. It excludes entire established roles in music where the primary contribution is direction, selection and shaping rather than manual sound generation.

A more consistent definition would be: a musician is someone who intentionally shapes musical outcomes and takes responsibility for them. The specific tools or level of manual control involved have never been a stable boundary in music history.

In the end, anyone who intentionally creates and shapes music can reasonably be considered a musician. The tools and level of direct control may differ, but that has never been a consistent boundary for musicianship.

Following that logic, it would also make little sense to argue that someone who has been writing and producing music for 30 years suddenly stops being a musician the moment they incorporate AI into their workflow. Changing tools does not erase authorship, experience or musical identity.

Musician or not?

[enroe]

Let's make a comparison in the world of sports.

In tennis, improvements in rackets made from graphite and carbon fiber have enabled players to hit with more power, control, and spin. Enhanced strings also add durability and performance, allowing players to be faster and more precise.
- This didn't make tennis players unemployed. Why? They embraced the new equipment.

In golf, advancements in club design, particularly with titanium and graphite materials, have allowed players to hit longer and more accurate shots. The development of high-tech golf balls also contributes to increased distance and control. Players have greatly benefited from these innovations, pushing the boundaries of performance.
- Did this make golfers unemployed? No. Why? They embraced the new equipment.

In table tennis, improvements in racket materials such as carbon fiber and high-quality rubbers have enhanced control, speed, and spin. These advancements allow players to execute faster and more precise shots. Players have taken advantage of these innovations to reach new levels of performance.
- Millions of chinese players had to quit because of this? I hardy think so. Why? They embraced the new equipment.

I can continue, but to conclude:

Just like athletes embrace new equipment to enhance their performance, producers and songwriters can use AI as a tool to create music. AI doesn’t replace the creative process, but opens up new possibilities, just as improved equipment helps athletes reach new heights.

No, no, we've been here before: You're confusing the "tool" with "creativity"!

As a "tool," AI can improve an equalizer, optimize compression, analyze the
frequency response, etc. There will likely be entirely new, amazing tools that,
with the help of specific AIs, will provide new possibilities.

That's NOT what's meant here. This is about AI as a "creative machine"—
like SUNO, Udio, and others. And they bring a completely new quality, thus
crossing a red line. They are incomparable to anything in the past because
they are truly "creative": They generate—via text prompt—complete songs
in a completely new way!

And that's a huge difference, practically the "atomic bomb" for human
culture. (The word "atomic bomb" was used by an OpenAI employee to
describe this new quality.)

This is what we're talking about here: this "atomic bomb." Many musicians —
I think most of them — don't understand what a neural, self-learning network
can really do — and what that means for the future. See here.

In your golf game analogy, it would be as if a robot were playing golf instead
of a human. And it would play much better than any human because its light
sensors allow it to calculate precisely how to hit the ball.

This completely changes the sport of golf - and makes humans essentially
superfluous.

[Tiles]

You’re still drawing a “red line” without actually justifying it.

Generative tools like Suno AI music generator or Udio AI music generator don’t remove creativity, they shift it. The user still defines direction, iterates, selects and shapes results. That’s a different workflow, not an absence of authorship. And yet again the important distinction between the AI tools is missing. It is not all Suno. And even in Suno it's not all throwing dices.

Calling this an “atomic bomb” is rhetoric, not an argument. Music has gone through multiple disruptive shifts before (recording, synthesis, sampling, DAWs), and none of them created a clear boundary where people suddenly stopped being musicians.

The “tool vs creative machine” distinction also doesn’t hold. Music tools have been generating material for decades (arpeggiators, drum machines, algorithmic composition). AI increases complexity, but not the underlying principle.

And the golf analogy doesn’t fit. Music isn’t a sport with fixed rules and measurable performance. It’s about shaping outcomes, not manually producing every detail.

Even if AI can generate full songs, humans still decide what to create, what to keep and what to release. That role hasn’t disappeared.

I still wait for the answer to my question that you dodged s bravely. Do you use equipment from after the sixties? Then you're not longer a musician by your own definition. So what is your battle here?

[enroe]

...

Now apply that to song writing. When I write a song, it pops out of me in some raw form, and then I hone it over time, often by performing it with a band. Those people bring themselves into it, and to some degree enhance and shape the music in a way that I couldn't really predict. Then I go into a studio and start recording and producing it. I'm making thousands of choices. Each one informed by years of my own experiences and my emotional connection to my experiences, all which have formed me. I'll use some technology to make parts of the job that I can't do possible, but it's all under my very direct control. It's not a fast process, but in the end, I have a deep feeling of accomplishment. That feeling fuels my next project, and the next after that, even if no one listens to my music, which trust me, they do not.

So since no one listens, why not just ask Sunos to make a song for me, based on my instructions? Because that's not making music, and I am a musician. If that's what you want to do, no one is stopping you, but if you think you're a musician, you're delusional.

Exactly - I agree with you! You describe the process and
procedure of songwriting very well.

And I can only emphasize this: In writing my own song, there is
so much feeling, experience, expression, communication - very
personal - yes, sometimes even sensitive and vulnerable, that it
is unmistakably connected to me as a person, that the song
practically represents "me as a person".

This is also why, conversely, song creation using text prompts —
or iterative text prompts, if you want to be a bit more complicated
— is empty, emotionless, and sterile. And this is despite the fact
that the AI-generated song is good in itself and hardly
distinguishable from a human-generated song by the audience.

Ultimately, we arrive at the general question:

Is the use of AI necessary for creating a song?

If you do that, the question arises: If you are a musician and a
human being, why do you let AI take over the most interesting
and essential aspects of composition? In doing so, you are giving
up on yourself. You are acting like an empty shell, creating
supposedly original songs with text prompts – not for personal
expression – but for money and fame?