I do music production. I’ve seen what AI can do, and for making a simple master of one of my demos AI is great. But everyone who has been there knows that the fun of making music comes from uniqueness. AI can’t do that, it can only riff on things known in its dataset.
I’m not worried at all, because this means the boring commercial music becomes less impressive, and unique and elaborate music becomes even more valuable. I’m cool with that; nothing of value is lost.
I get why people are freaked out, but most of the tangible issues come from greedy managers with a pinch of dunning-kruger, trying to replace workers. And that’s got almost nothing to do with AI and everything to do with assholes. There will be consequences, but people have to stick out this short-term chaos. And that’s the only thing I have empathy for in this.
Well that and unlicensed training because governments can’t get their shit together to enforce ethical guidelines for AI training & data collection practices.
Are you actually using an AI mastering service? Ozone’s “AI”, which I use for mastering seems to be anything but.
I’ve drawn a line at AI should not make it into the final product (I want to be able to manually tweak everything) I’m fine with whatever Ozone does when configuring itself.
I think AI is also a great tool to get people into hobbies they will want to do themselves.
A lot of things - including music production - can seem daunting for a beginner, and for all the resources available, none can replace a teacher who sits next to you and guides you through the basics.
And fortunately I’m seeing more and more people using AI not in place of a professional but to become one, utilising the LLM to create personalised teacher agents. And LLMs are arguably much better at “teach me to do this” rather than “go and do this”.
I say this as a software developer and producer/dj hobbyist with a solid understanding of AI and it’s pitfalls:
I genuinely have not had any kind of positive experience with LLMs and music writing. I hate Suno but it is at least trained on music. It can kinda approximate the target.
But asking an LLM about music production has yielded so much garbage (likely because it has been trained on garbage- the music producer influencer sphere is notoriously full of bad and anti-creative advice).
It’s generally ok for music theory, but ask for say, a sound design walkthrough and you’re gonna be left scratching your head. Ask it for arrangement advice and you’re gonna get incredibly generic advice. It hallucinates workflows and plugins between DAWs (i.e. suggesting using an Ableton plugin in FL). And like sure, I can
So, I don’t think we’ve synthesized the requirements for getting a music production teacher out of llms yet. And to be clear, I hate where ai is taking the world and that it’s forcing questions about human authorship, but I’m also not a “everything it outputs is wrong” person. Just this specific use case isn’t there yet.
I wish your opinion was shared more.
I do music production. I’ve seen what AI can do, and for making a simple master of one of my demos AI is great. But everyone who has been there knows that the fun of making music comes from uniqueness. AI can’t do that, it can only riff on things known in its dataset.
I’m not worried at all, because this means the boring commercial music becomes less impressive, and unique and elaborate music becomes even more valuable. I’m cool with that; nothing of value is lost.
I get why people are freaked out, but most of the tangible issues come from greedy managers with a pinch of dunning-kruger, trying to replace workers. And that’s got almost nothing to do with AI and everything to do with assholes. There will be consequences, but people have to stick out this short-term chaos. And that’s the only thing I have empathy for in this.
Well that and unlicensed training because governments can’t get their shit together to enforce ethical guidelines for AI training & data collection practices.
Are you actually using an AI mastering service? Ozone’s “AI”, which I use for mastering seems to be anything but.
I’ve drawn a line at AI should not make it into the final product (I want to be able to manually tweak everything) I’m fine with whatever Ozone does when configuring itself.
I think AI is also a great tool to get people into hobbies they will want to do themselves.
A lot of things - including music production - can seem daunting for a beginner, and for all the resources available, none can replace a teacher who sits next to you and guides you through the basics.
And fortunately I’m seeing more and more people using AI not in place of a professional but to become one, utilising the LLM to create personalised teacher agents. And LLMs are arguably much better at “teach me to do this” rather than “go and do this”.
I say this as a software developer and producer/dj hobbyist with a solid understanding of AI and it’s pitfalls:
I genuinely have not had any kind of positive experience with LLMs and music writing. I hate Suno but it is at least trained on music. It can kinda approximate the target.
But asking an LLM about music production has yielded so much garbage (likely because it has been trained on garbage- the music producer influencer sphere is notoriously full of bad and anti-creative advice).
It’s generally ok for music theory, but ask for say, a sound design walkthrough and you’re gonna be left scratching your head. Ask it for arrangement advice and you’re gonna get incredibly generic advice. It hallucinates workflows and plugins between DAWs (i.e. suggesting using an Ableton plugin in FL). And like sure, I can
So, I don’t think we’ve synthesized the requirements for getting a music production teacher out of llms yet. And to be clear, I hate where ai is taking the world and that it’s forcing questions about human authorship, but I’m also not a “everything it outputs is wrong” person. Just this specific use case isn’t there yet.