AI + Music --> What will the future look like?

[superturrican]

AI music is a fun diversion if you need a quick joke song to send to someone or if you need musical wallpaper for something corporate. It's not art, because no human imagination was involved (unless you think search terms are also creative works), but then muzak isn't art either, so it has a use case. Not sure if you can copyright an AI prompt, is that something that's being tested in court?

[Kalamata Kid]

Keep in mind that this is baby AI.

AI vs Mozart: Can YOU tell the difference?

[Bunny_boy]

When he says that it's only an excerpt, I think that's key. Anything to do with imitation works best in short bursts. It would be interesting to hear (or be subject to) a full composition, to see if any themes, etc are developed. Fun video as well

[vurt]

AI vs Mozart: Can YOU tell the difference?

the ai isn't wearing a powdered wig.

[Kalamata Kid]

A full composition is more difficult for both humans and Ai.
Perhaps in a year or two a they can make the same test performing a full composition.

[sineforge]

I think for us producers, we need to be unique. LLMs are rehashing old patterns, they have a hard time improvising, and if they do this, it sounds stale, silly and otherwise not cool.

The other thing I think is that LLMs lack are a human personality, so you need to be more upfront about your background, why you are doing music and so forth, so the listeners have a connection to a human and not an algorithm.

I think one example is the connection to art and artists -- for example last night I read stories about Miro the painter. I would not read real life stories about Claude's LLM algorithm.

I struggled with depression for a long time at the beginning of the whole AI music hype (feeling useless), but eventually, while working on it, I came to the same type of thoughts.
I hope listeners feel it too, though

[eassae]

[Aloysius]

All what?

[lobanov]

1.

We should always know who the real composer and real performer are. And we should want to know it. That's the main rule. If it's Suno, it's Suno. The important thing is that it's explicitly written.

2.

And this has extreme consequences: Both individuals and music companies
are flooding Spotify and all social media platforms like Facebook and TikTok
with a constant stream of new, great songs. And the kids and teenagers in
all countries are totally into it.

The big companies and those dominating Spotify have discovered:
With AI creators, they can generate songs in no time, upload them in droves,
and even make money. Companies are even using AI to design live groups,
partly human, partly avatar, which are presented as live performers and deliver
fantastic, breathtaking, even superhuman live performances. TV and internet
platforms will be flooded with them — teenagers are intoxicated by them and
click on them.

AI generated music lacks commercial potential. I mean cases like Beatles, Michael Jackson etc. Thousands of people pollute social media and streaming services with hundreds of thousands of mediocre and similar to each other 'great' songs... No one wins. The more people involved, the less profit. But a lot of people participate because it's affordable and easy.

You have to do significantly more to succeed. Maybe you have to generate a thousand songs and pick one out of them. Or choose from them a few fragments that are really interesting.

AI is designed to generate huge amounts of music in a short amount of time. You can generate music faster than you can listen to it. And such a ‘product’ is only good for generating a 'vague' music stream. Background music-like noise. That's the nature of this technology.

[Aloysius]

I want to generate songs by my favourite band. The leading members are dead. Is it possible to take a few tracks and generate a new one in a similar style just to listen to? There are guitars. Maybe AI would use a crappy GM preset called guitar.

[dexrow]

I think the Iranian independence song they are distributing on one of the iran channels is ai.. sorta a strange expierence

[enroe]

There's always multiple trends happening at the same time.
It's been fascinating to see how artists have responded to the new set of technological limitations that they inherited, the limitations being generative. But the fact that the inherited limitations were new was a result of experiments aimed at overcoming the limitations of technology, which is happening constantly.

So I'm guessing that there's not really any limit in terms of reducing the friction between traditional DAW tools, and AI tools. Currently they exist fairly independently. But every time a limit is imposed in terms of what's viable in integrating them, users will want those limits to be superseded. So slowly there'll be a move towards a combination of full user control, along with the deepest possible AI automation, with the user having as much choice as possible of when to use either.

And perhaps it's depressing that the convenience and sophistication of the AI tools will erode the motivation towards traditional human user input.
I don't see why it should be limited to EQ and compression, surely synthesis will move in the direction of having prompt-based input, with users just asking for 'more growly', 'more smooth', 'more angry' etc.

During a recent social gathering people were playing favourite music from different decades, and quite a bit of it I'd forgotten existed and forgotten how good it was. The uniqueness of the circumstances in which that music was made can never be repeated, and perhaps we're moving towards a time when there's just won't be any motivation or reward for the kind of artistic experimentation which was for a long time a durable part of culture. The trade-off will be a level of immediate plasticity which can be used for ephemeral purposes, e.g. if you make a podcast covering topical events, you can with almost no effort incorporate AI soundtracks which combine all the topical elements.

Then again, perhaps the market won't develop that way. If the main driver towards AI integration in DAWs is the speed at which professional musicians can produce work, with the net impact being that a smaller number of producers get the majority of the work, and if the majority of hobbyists don't want AI tools, then perhaps there won't be such an economic incentive to accommodate that increasingly smaller number of professionals.

Hmmm... yes, I've now read your post three times.

And I think almost all of your thoughts relate to Chapter 1:

1. AI tools + AI plugins

The use of AI tools is exactly as you describe: an almost linear expansion
and improvement of the tools. This will apply to equalizers, compressors,
and other tools. Perhaps new AI-supported overview and organization
functions for the DAW will also be offered. All of these will, of course, make
the composer's or mixer's work with the DAW easier and better.

The more interesting chapter, however, is (2):

2. AI Song Composition

And here – in my opinion – things are quite different. Because AI introduces
a completely new quality. It's no longer "simple data processing." It's not even
"complex data processing." No – AI consists of "neural networks" that learn
almost like biological networks, much like a small child. After all, we also have
many, many neurons. The "neural networks" are very similarly composed of
such neural connection points. By linking billions of these points, such
systems learn – much like a small child, only 100,000 times faster.

We are now seeing the first signs of such AI systems in the field of language
(language models), but also already in the field of music composition. You
describe the advantages as "a level of immediate plasticity" that "can be
used for ephemeral purposes." But AI systems will be applied to many more
areas.

In just a few years, there will be very comprehensive AI systems. And that
brings us to point (3):

3. AI song composition - in 4 years and beyond

This won't simply be an improvement to "any old tools." No, this will
disruptively change the music world. AI will completely take over composing
and mixing.

So not "just any" background music, but iconic hit songs, composed of
familiar elements, yet new. Songs that will conquer popular platforms like
Spotify, dominate TV and radio stations, and even increase the "hit rate."
For teenagers and listeners, this will be a new, perhaps great, era. However,
human creativity—and thus the individual musician—will no longer play a
role in this game.

The question for me is:

• How will we as musicians react when AI creates complete, and truly good songs?

• What significance will composing music even have if we give free rein to AI?

• How should society react to this – in an ideal world?

[enroe]

The immediate threat is when the current AI bubble pops. Open AI has no realistic plan ...

Yes, the stock market – that's a whole world of its own. There are usually
many surprises, and the real winners only become clear after many years.
Just like with the internet in the 90s and 2000s.

... AI is making the effort to learn to play an instrument, sing, or write songs outdated and old fashioned. You can get "good enough" instant results from typing in a one sentence prompt to Udo/Suno so why bother? ...

Mmm, yes, but don't I hear a sense of powerlessness, a feeling of futility?
(I have that too, though.)

We – you, me, all musicians who truly compose – have explored the vast
creative realm of human inventiveness and accomplished incredible things
there – and continue to do so. It has become an important part of our identity.
And is all of that supposed to end now? Because AI, Suno, and the like can
do all of this at a moment's notice, at a "prompt" – and will do it even more
perfectly in the future?

Should we give up all of that—our own creativity and identity—and shrug
and say, "That's just progress"?

[eassae]

Should we give up all of that—our own creativity and identity—and shrug
and say, "That's just progress"?

I for one enjoy the process of making music. I like all of it, the sound design, the writing, mixing, and although I'm not the best pianist, I do like playing. Sometimes finishing an album is tedious because going through the details I sometimes don't know if I'm making something better or worse, but even that is worth it in the end, and I definitely wouldn't trust the process to a machine. I guess even if I did trust the machine would do a good job, I wouldn't feel good about using it as self expression in an artistic manner is something sacrosanct to me.

I don't think I'll ever understand why a person would want to bypass the process of creating something meaningful, whether it reaches an audience or not. I don't have to though, but I do think a mechanistic culture that is willing to skip all steps to a result is harder to live in…for me anyway.

[chagzuki]

So not "just any" background music, but iconic hit songs, composed of
familiar elements, yet new. Songs that will conquer popular platforms like
Spotify, dominate TV and radio stations, and even increase the "hit rate."
For teenagers and listeners, this will be a new, perhaps great, era. However,
human creativity—and thus the individual musician—will no longer play a
role in this game.

The question for me is:

• How will we as musicians react when AI creates complete, and truly good songs?

• What significance will composing music even have if we give free rein to AI?

• How should society react to this – in an ideal world?

Depending on the requirements dictated by your musical tastes, AI may already be at a sufficient level of sophistication. For me, the majority of 'iconic hit songs' are shit, but then I'm more interested in classical music.
Music became less important to youth culture a long time ago, really since social media. Prior to the kind of instant communication that we have now, the need for social predictability and a shared set of symbols/norms lent music a special status.
If there's still a need for music in the pop template in the future, it's quite feasible that it gets generated automatically as a result of one's social media interactions, i.e. it'll just pop up in your feed etc. Or you can just have an endless stream of new songs generated on the fly according to some choices.
For years now corporate pop music has been recycling every permutation of euphemism for sex/romance, for every audience. Oh, so hot, so cold, ice cold, fever, you left me, I want you back, I don't want you back, I can get all the bitches I want with my gangsta factor, my T & A are the best etc. May as well have AI take over there.

However, people are weird... there are different markets. Teenagers and 'idol' type bands, that seems to be an enduring phenomenon. I guess there's a formative window around the age of 12-16 where it's common to latch onto examples of... er, people.
People who go to clubs, apparently, seem to want to imagine that there's some kind of special DJ figure who's doing more than pressing the play button.
People go to see 'live' bands who are miming.... all kinds of weird nonsense.
So who knows which of those things will somehow endure in an era where AI is ubiquitous?

Presumably some people will still want to see real people playing real instruments, jazz, rock, classical etc., but I don't imagine that will be a huge market.

Can music still be a kind of sonic imprint, an expression of the times as lived by the artist? Maybe, but there's certainly less of a need for that sort of thing. People can simply talk about their experiences now, create podcasts etc., speak instantly with anyone anywhere.

I think it's more likely that the extra plasticity will allow for ever-evolving background music in games, and interactive environments will dominate entertainment. And also learning, i.e. virtual learning environments.