Poll : AI in sound and computer-assisted music

[Arnaud67]

Hello everyone,

I'm a musician and researcher at the University of Strasbourg and I'm doing some research to find out what role generative AIs play in sound and computer-assisted music, today and tomorrow.

It would be a great help if you agreed to answer this poll and comment it.

Strong opinions are welcome.

Many thanks for your invaluable contribution.

To the moderators: thank you in advance for moving this message if it is not in the right place, with my apologies for the inconvenience.

Best regards,
Arnaud

[jancivil]

The modifier 'generative' jumped out at me as I read over what you wrote a second time.

So, I am not against machine learning for technical things, in mix/mastering processes for instance, but I generate my own music, so that's where my "there's no reason to use it" answer locates.

I do at times set up a series of processes quite beyond my direct control and impossible to predict because of complexity, and use technology in ways it wasn't intended to do (for instance, polyphonic audio to MIDI when it's only made to deal with monophonic sources) which generates things I won't have thought of at all, but the germ was my thinking and the outcomes have a certain integrity with my idea. So in a sense the computer assisted but it didn't have to do any thought or really even a semblance of.

[waxtrax]

I see the usage of AI in music as no different than when music became electrified during the last century, in all of its different forms. AI is a tool that can be used to accomplish different tasks in different ways, much like an array of guitar pedals or traditional rack-mounted signal processors are tools.

Every time something new comes along, people will complain and say it's not "true" music, and how can people listen to that garbage anyway? I find it hilarious that people still have this argument. What's "real" and "good" to one person can be completely irrelevant to another person. Art is in the eye of the beholder, regardless of the tools used to make it or the media on which it is presented.

[Shabdahbriah]

I picked "there's no reason to use it", for similar reasons to jancivil. I do a lot of software modular, and eurorack stuff that is at times stupidly complex, because I love the build-up and 'happy accidents'/serendipity. I've quite a arsenal of randomness available. That said.
I do have a few apps like Synplant and Spectralayers, that use AI assisted analysis, but I have no desire to use (say) an on line app, to generate a track, or any other process, as I've been recording/editing/mixing for 50+years, and enjoy those disciplines/procedures. I may use one (someday) for voice-over dialogue in youtube tutorials, however.

[lotus2035]

The past few decades the entire music industry gradually shifted from being run by people who loved music or at least respected music and musicians to being gutted and hollowed out by business minded types who only cared about their bottom line. Thanks to A.I. a similar thing could now happen to music itself.
Music created by actual musicians will be side-lined and crowded out by the millions of tone deaf, talentless, ass-hats who will aggressively gut the entire historical musical catalog of humanity so as to train their 'database' and upload the results everywhere with the intention of creating new passive income streams to go along with their other investments.

Ways to stop it might be to ban the harvesting of musical data entirely or make the results of it 'for personal use only' with no money making possibilities allowed. But if it continues to improve how will you know which is the original and which is the copy? You won't.
And you might get sued by some A.I. music conglomerate because your music sounds exactly like the music they made from your music.
But hey we'll still have our G.A.S. to keep us busy and software companies will keep selling us stuff.

Now, I had my eye on something in the marketplace.

[liqih]

Like everything does, "Find new ideas and inspiration" got my vote, but it's so obvious, isn't it?

Also I voted for "Achieve results at lower cost", not because I use it like that, just I think that's what AI in sound and computer-assisted music can currently offer.
I mean the quality it's not there (yet?). I tried a few AI tools to generate sounds or songs or parts and I still felt some randomness in their output.

Anyway, "at lower cost" is also debatable, given huge energy and/or data consumption needed by an AI machine.
Plus the downgrading of human actions to prompts writing, one could say. This line could be seen as the moving of the creative activity from musician to manager.

[Trancit]

Everybody dreaming wet dreams about wannabe "AI" should know this video and especially what it means for copyright and the ongoing stealing of real people´s work:

[swilow11]

Find new ideas is valid imo. Or just having higher level concepts explained simply (but this application is much broader than just music).

So far, I really haven't used AI in music production and am certainly not excited by the concept- quite the opposite. I've asked ChatGPT to provide formulae to use in Serums wavetable editor (help me make a vowel/formant table) and the results were interesting. Have asked for sound design ideas (how to make thunder?) and the results were pretty bad. I can see AI helping with rhe complexity of something like a complex Reaktor instrument, but less interested in the notion of AI actually creating music.

I predict it will become 'better' than the best human producers and soon, but I think it's always going to miss one vital component for music appreciation- that being virtuosity or technical chops. When I hear a complex drum groove/glitch sequence/guitar shredding thing, on one level I'm just loving the sound, but alongside that is admiration for the technical abilities behind the thing. Knowing that it was created by AI will just erase that aspect of appreciation. I get this with the hype realistic AI art- if I know it's AI, the experience is somewhat cheapened.

[trusampler]

AI will find it's best use cases among the current lackluster audio browsers. Key detection / Bpm detection are area's that most music producers won't argue that need real improvement and you won't face much push back.
Good luck with your research.

[BertKoor]

Paraphrased, read something like that somewhere, can't find the source.

"While it is easy to create something, and anything can be created, it has become extremely difficult to create something of value or meaning"


Last option: I don't need it. What we have now are stochastic parrots. Given plenty examples, it can generate more of the same. That is not creativity.

Note: bpm, note & key detection are relatively simple problems. No A.I. required, and it is not 'generative'.

[jonljacobi]

Just from the level of personal satisfaction, I see no way to get joy from pressing a button and having your computer spit out a composition. As a source for study materials, I can see it as a tool. However, I'm aware that some people want the money and could give a shit about personal growth and reward.

[vurt]

i chose no reason to use it.
i mean for me, im not suggesting no one else has a reason, just for me personally, i enjoy the whole process of it, even (or perhaps especially) when it's going wrong, that for me is where my brain really engages and i have to make it work, and fast! as im generally recording modular live, it's that bringing it back to a more tame sound, before letting it ride again, that i look for in my sessions.

[EnGee]

I don't find any fun in using AI in music or arts. It is actually a burden! Suppose I depend on AI Drums in Logic, then even if I have all the controls but in the end I will be 'slave' to the result that AI produce. It is much better that I do it by myself calculating carefully (and learning) how to lay the drums according to my song. However simple, I find the result more interesting like home made cooking
So, of course I chose "no reason to use it".

[osiris]

You know I have all the AI stuff and it's generally useless or horribly hobbled. You need to add an option 'Will pay for AI that can actually do something', since Google is now doing drug pusher type deals for its new AI. (First month free, then $19.99 per month)

[tommy2tone]

I recognize its potential value in finding new ideas to explore but on the flip-side, I've never heard any "AI-generated music" that actually emotionally moved me. AI lacks soul.