Here's What Gemini Thinks About the Whole Creative AI Argument

[BONES]

Anyway, here is an example that is like what you are currently doing. The vocals are all Ace Studio and the rest is regular DAW and live playing.

I love the bass parts in this - great sounds, interesting change-ups. It's just a pity that whoever decided to call this "Cyberpunk" totally forgot about the "punk" aspect of it. It's way too lightweight.

So what if the AI comes up with a sound I don't like? See, with a sample library the auditioning of sounds is seconds. With AI, it would be hit and miss and still potentially take a long time to come up with something I like.

OTOH, if you heard a sound in another song that you really liked, that your knew would be perfect for something you were working on, you could play it to the AI and it would reproduce it in seconds, which would save you a tonne of time. So in any given situation, you'd use the tool that you thought would get the job done the fastest, so you could get on with your creative work. I have to say, though, that I find previewing patches soul-destroying and so much of the time it takes me out of the moment. It's something I tend to do later in the process, once everything is sort of working. The ability to do it on the spot could be very handy.

What's this going to cost me?

That's something we don't yet know. A subscription model seems most likely to me, which I can live with. Pay for it while you need it, then unsubscribe (just like Netflix). That would work really well for me, as I tend to use the shit out of a new plugin for a few months, then move on to some newer, shinier thing. We did it with Output's Arcade - subscribed while we were working on an album and once it was all finished, I just rendered those tracks to audio files so we could use them live and cancelled the subscription.

Yeah, I got no problems with vocal AI. Thanks to it, I will never have to sing on a recording again. Just sing well enough to give the AI something to work with.

It's funny (strange) because vocals are the absolute last thing I'd give up to AI (although I have already, in one song on our 6th album). Doing vocals is really the only aspect of all of this that I truly enjoy. I just spent a few minutes trying to match it with Peter Murphy on his new album, Silver Shade. There is nothing more joyous than belting it out at the top of my lungs. It's where the real connection with the music comes from for me.

Again. what if I don't like the sound the AI comes up with?

Give it some feedback and make it try again. If it can generate 6 versions of a song in a few minutes, I imagine creating a single instrument sound would only take a few seconds.

[wagtunes]

@Bones. All good points. If AI ever did get to the point where it was no different from using a plugin to pick a sound or effect or whatever, of course I would use it. I am not against the concept of AI. I don't even care if people want to push a button and make a song and be done with it. If they're happy doing it, power to them. It's just not for me. I would get no enjoyment out of doing that.

[jancivil]

I refer to that as "chord progression". Chord structure to me are more about inversions and spacing and whether the chords are open or closed. Because I've been exploring more into Bartok Axis theory along with different ethnic scalings, where I thought it might break me into new territory, I've found I've been doing much of it all along. And yes, the study is still helping to expand further.
I know that's a bit off topic.
More on topic perhaps...
If you tell AI to make it in the style of Marvin Gaye, how quick will the family sue you?

Oh for the love of God, semantics. Are you REALLY going to be this anal about this? Chord structure, chord progression, chord choices, key signature, key changes, it's all the same damn can of corn.

all you show is you don't understand what was said about the topic. It's a specific idea, people have them. You are happy doing less, and apparently being an incurious sort. Instead of finding out what the idea is you dismiss it and make this stupid reduction all harmonic thought is the same. To you; ok, so you be you. But why be a dick for no good reason at all?

Bartók's Axis Theory, developed by Ernő Lendvai, analyzes Béla Bartók’s music by organizing the 12 chromatic tones into three functional axes—tonic, dominant, and subdominant. Each axis connects four notes, separated by minor thirds, where opposing "poles" (e.g., C and F#) act as substitutes. It expands traditional harmony by enabling functional substitution across tritone relationships.

It's a truly obvious logical fallacy what you did there making things equivalent just because they're part of the same category. You literally have no idea what you're even saying.

[wagtunes]

I refer to that as "chord progression". Chord structure to me are more about inversions and spacing and whether the chords are open or closed. Because I've been exploring more into Bartok Axis theory along with different ethnic scalings, where I thought it might break me into new territory, I've found I've been doing much of it all along. And yes, the study is still helping to expand further.
I know that's a bit off topic.
More on topic perhaps...
If you tell AI to make it in the style of Marvin Gaye, how quick will the family sue you?

Oh for the love of God, semantics. Are you REALLY going to be this anal about this? Chord structure, chord progression, chord choices, key signature, key changes, it's all the same damn can of corn.

all you show is you don't understand what was said about the topic. It's a specific idea, people have them. You are happy doing less. And you're being a dick for no good reason at all.

I'm happy doing less because I use AI vocals because I can't sing for shit? Are you f**king kidding me?

I have NEVER used AI to write a song EVER.

So how am I happy doing less?

You're the one being a dick.

[jancivil]

f**king dullsville, mang

[jancivil]

Gemini doesn't think anything about anything. It does have a capability to curate what people have said, and it's trained to do quite a lot in this regard evidently. I asked Google 'can Suno be said to understand' two different times, and it provided knowledgable answers. from places like Reddit. (The second answer sounded like me talking.). It has no awareness and no mind, if it appears to have an opinion it''s replcating what it found. Gemini 'said' certain things there that are close to exact replicas of statements I've seen here. I'd check sources, it should provide them.

[jancivil]

"I - iV - V - I has been a staple of western music for centuries... but Bartok wasnt a fan." and so we see a different functional paradigm. For reasons.

Which isn't terrifically surprising if you know his music. I see that and I'm curious. I'm glad I looked it up, thank you BBFG#.

I'm not at all likely to use it but its there to maybe understand Bartok's thought in certain tnings.

[jancivil]

within the limitation of 12 equidistant tones the ordering of various combinations of in all the possible distances in voicing can be so expansive one wonders if it isn't infinite. Here is an example of a specific idea, unlike or diredtly comparable to ideas from vanilla diatonic constructions:

"symmetrical octatonic scale; restricted to three possibilities, as transposition is limited by that ordering. aka the diminished scale. (0, 1 represents the set beginning as if C, Db; 0, 2 as if C, D; 1, 2 as if D, Eb)

---

octatonic examples.png

so this one idea can be explored endlessly and things found that one hasn't found before that are unique to the material itself. 8 notes. Music is not limited except by one's imagination or prejudices.

[BONES]

"I - iV - V - I has been a staple of western music for centuries... but Bartok wasnt a fan." and so we see a different functional paradigm. For reasons.

The thing is, if you don't know any of this shit, and it means less than nothing to me, then you aren't bound by any of it, you're free to use whatever notes you think work together in any given situation. It's instinctive, not constrained by rules and conventions. If it sounds good, it is good and that's all that matters.

[zerocrossing]

What does zerocrossing think about AI?

Based on discussions on KVR Audio and Gearspace, zerocrossing (a long-time, prolific KVR member) generally views AI in music production with a mix of cynical pragmatism, skepticism about its "artistic" value, and an acknowledgment that it is an unstoppable technological shift.
His views, often expressed in threads about AI adoption, can be summarized as follows:
AI as a "Random Inspiration Machine" or "Tool": Zerocrossing views current AI tools not as true creators, but as advanced, albeit sometimes useful, assistants. He has noted that while AI can be great for overcoming writer's block or generating ideas, it is not fundamentally creative.
Skepticism of "Writer's Block" and "Prompt Jockeys": He often challenges the narrative that AI is needed for productivity. He has implied that if someone constantly relies on AI to overcome writer's block, they might be in the wrong hobby, stating, "most composers and musicians I know almost never have writers block".
Focus on the "Human" Element: He tends to emphasize that the value of music is in the human experience of creating it—the "doing," the "making," and the emotional, often messy process, which AI cannot replicate.
The "Super-Genius Parrot": He has described AI as a "super-genius parrot" that can remix and regurgitate information, but it lacks true consciousness or understanding, arguing that "intelligence" in AI is often marketing hype.
Pragmatic Adaptation: Despite his criticisms of AI art's lack of soul, he is not completely opposed to using AI tools. He has expressed interest in how AI can be integrated into an "indie approach" to music, noting that some AI-generated music can be "listenable" and "interesting".
"Music had a one night stand with sound design.....And the condom broke." This humorous quote from a recent KVR post (Feb 28, 2026) highlights his view that AI has irrevocably changed the landscape of music production.
Gearspace
Gearspace
+5
In summary, zerocrossing is not an anti-tech purist, but a skeptical artist who believes AI is a "soulless" tool that cannot replace the emotional, and, "alchemical" process of human creation, even if it is a useful assistant.

Most of this is not true, except that I'm a prolific poster on KVR.

[BONES]

What sort of giant narcissist asks a question like that?

[KBSoundSmith]

What does zerocrossing think about AI?

Based on discussions on KVR Audio and Gearspace, zerocrossing (a long-time, prolific KVR member) generally views AI in music production with a mix of cynical pragmatism, skepticism about its "artistic" value, and an acknowledgment that it is an unstoppable technological shift.
His views, often expressed in threads about AI adoption, can be summarized as follows:
AI as a "Random Inspiration Machine" or "Tool": Zerocrossing views current AI tools not as true creators, but as advanced, albeit sometimes useful, assistants. He has noted that while AI can be great for overcoming writer's block or generating ideas, it is not fundamentally creative.
Skepticism of "Writer's Block" and "Prompt Jockeys": He often challenges the narrative that AI is needed for productivity. He has implied that if someone constantly relies on AI to overcome writer's block, they might be in the wrong hobby, stating, "most composers and musicians I know almost never have writers block".
Focus on the "Human" Element: He tends to emphasize that the value of music is in the human experience of creating it—the "doing," the "making," and the emotional, often messy process, which AI cannot replicate.
The "Super-Genius Parrot": He has described AI as a "super-genius parrot" that can remix and regurgitate information, but it lacks true consciousness or understanding, arguing that "intelligence" in AI is often marketing hype.
Pragmatic Adaptation: Despite his criticisms of AI art's lack of soul, he is not completely opposed to using AI tools. He has expressed interest in how AI can be integrated into an "indie approach" to music, noting that some AI-generated music can be "listenable" and "interesting".
"Music had a one night stand with sound design.....And the condom broke." This humorous quote from a recent KVR post (Feb 28, 2026) highlights his view that AI has irrevocably changed the landscape of music production.
Gearspace
Gearspace
+5
In summary, zerocrossing is not an anti-tech purist, but a skeptical artist who believes AI is a "soulless" tool that cannot replace the emotional, and, "alchemical" process of human creation, even if it is a useful assistant.

Most of this is not true, except that I'm a prolific poster on KVR.

I'm a bit of a late adopter when it comes to trying AI, but I've been experimenting with Claude, ChatGPT, and Gemini a lot recently. They each have distinct strengths and weaknesses (some of which I haven't fully explored for pay-wall reasons, but the implied use case is clear both from the experiments I've done and the pay-walls themselves). Some of my deeper experiments have gone into rabbit-holes which the engines themselves have described as "atypical", "black swan events" and so on. Interesting, but I don't have an intention of discussing those.

Rather, I'd offer a comparison. I'm reminded of software vs hardware synthesizer debates. While the debates of the early days of VSTs were before my time, so to speak, looking back to them and comparing, I see strong parallels to AI models today. Early softsynths showed promise, could give decent output (or trash, haha), but "analog purists" would have none of it (and often still don't, haha). Now I'm not indicating that AI is "the way of the future" -- I'm not one for jumping on hype trains, especially when there's billions of dollars behind it -- but based on my experiments, I see (for targeted use cases, and utitlizing specific models and setups for specific use cases) the promise in these models that some saw in the early days of softsynths.

Regarding Gemini specifically, it naturally wants to look at large volumes of data, do a deep analysis, and then provide a "synthesis" of that data. The free tier is locked to small volume context and operates quickly -- pretty much the opposite of its intended use case. So the free version is prone to hallucinations and unwarranted grand conclusions from small amounts of data. Althought FWIW, for some of my use cases, I actually find this a charming "lo-fi" glitch which has some interesting applications.

Anyway, just a point of interest that flashed into mind and I thought I'd share.

[VariKusBrainZ]

AI cant and doesnt ‘Think’ and I concur with the AI synopsis of Zerocrossing’s assessment of it.

[zerocrossing]

What does zerocrossing think about AI?

Based on discussions on KVR Audio and Gearspace, zerocrossing (a long-time, prolific KVR member) generally views AI in music production with a mix of cynical pragmatism, skepticism about its "artistic" value, and an acknowledgment that it is an unstoppable technological shift.
His views, often expressed in threads about AI adoption, can be summarized as follows:
AI as a "Random Inspiration Machine" or "Tool": Zerocrossing views current AI tools not as true creators, but as advanced, albeit sometimes useful, assistants. He has noted that while AI can be great for overcoming writer's block or generating ideas, it is not fundamentally creative.
Skepticism of "Writer's Block" and "Prompt Jockeys": He often challenges the narrative that AI is needed for productivity. He has implied that if someone constantly relies on AI to overcome writer's block, they might be in the wrong hobby, stating, "most composers and musicians I know almost never have writers block".
Focus on the "Human" Element: He tends to emphasize that the value of music is in the human experience of creating it—the "doing," the "making," and the emotional, often messy process, which AI cannot replicate.
The "Super-Genius Parrot": He has described AI as a "super-genius parrot" that can remix and regurgitate information, but it lacks true consciousness or understanding, arguing that "intelligence" in AI is often marketing hype.
Pragmatic Adaptation: Despite his criticisms of AI art's lack of soul, he is not completely opposed to using AI tools. He has expressed interest in how AI can be integrated into an "indie approach" to music, noting that some AI-generated music can be "listenable" and "interesting".
"Music had a one night stand with sound design.....And the condom broke." This humorous quote from a recent KVR post (Feb 28, 2026) highlights his view that AI has irrevocably changed the landscape of music production.
Gearspace
Gearspace
+5
In summary, zerocrossing is not an anti-tech purist, but a skeptical artist who believes AI is a "soulless" tool that cannot replace the emotional, and, "alchemical" process of human creation, even if it is a useful assistant.

Most of this is not true, except that I'm a prolific poster on KVR.

I'm a bit of a late adopter when it comes to trying AI, but I've been experimenting with Claude, ChatGPT, and Gemini a lot recently. They each have distinct strengths and weaknesses (some of which I haven't fully explored for pay-wall reasons, but the implied use case is clear both from the experiments I've done and the pay-walls themselves). Some of my deeper experiments have gone into rabbit-holes which the engines themselves have described as "atypical", "black swan events" and so on. Interesting, but I don't have an intention of discussing those.

Rather, I'd offer a comparison. I'm reminded of software vs hardware synthesizer debates. While the debates of the early days of VSTs were before my time, so to speak, looking back to them and comparing, I see strong parallels to AI models today. Early softsynths showed promise, could give decent output (or trash, haha), but "analog purists" would have none of it (and often still don't, haha). Now I'm not indicating that AI is "the way of the future" -- I'm not one for jumping on hype trains, especially when there's billions of dollars behind it -- but based on my experiments, I see (for targeted use cases, and utitlizing specific models and setups for specific use cases) the promise in these models that some saw in the early days of softsynths.

Regarding Gemini specifically, it naturally wants to look at large volumes of data, do a deep analysis, and then provide a "synthesis" of that data. The free tier is locked to small volume context and operates quickly -- pretty much the opposite of its intended use case. So the free version is prone to hallucinations and unwarranted grand conclusions from small amounts of data. Althought FWIW, for some of my use cases, I actually find this a charming "lo-fi" glitch which has some interesting applications.

Anyway, just a point of interest that flashed into mind and I thought I'd share.

So, I read that as, "if you want an actual answer to a question, you need to pay for it." Of course, this isn't really talked about enough, IMO. I donate $100 to Wikipedia every Christmas. For that, I can go and do a search on nearly every topic, and get good information, or I could not donate anything and go to Wikipedia and get the same good information. So for $240 a year, I could subscribe to the better version of ChatGPT that has some "advanced reasoning," (not really) and get good information... maybe. But it will make me an image of cats dancing at a night club, so that's something. I could go the free version, and get bad information! How quaint!

The new version of ChatGPT is 25% better than the previous version. It also uses 4 times as much energy to get this 25% increase. It's out of training data, having looked at everything on the web. Because a lot of the web is now AI generated, it is eating its own sh!t at an alarming rate. So, unless there is some magical breakthrough, we've already hit a plateau. This is very different than plugin instruments. Most of the issues with early plugins were that they had to work in real time with a reasonable amount of voices, so shortcuts were taken so CPUs didn't melt. CPUs got much faster, added more cores, so developers could take fewer shortcuts and get excellent results. This is very different than one AI is doing, which is throwing a sh!tton of off-site processing to make the possibly correct information happen.

[KBSoundSmith]

AI cant and doesnt ‘Think’ and I concur with the AI synopsis of Zerocrossing’s assessment of it.

Correct, it doesn't think, it works on probability and pattern recognition. Yes it was inaccurate for zerocrossing's usage. Did I say anything contrary to that? No I did not.