[raven105x]

To be fair, this is no different than what's done by a human artist. Virtually every song composed today is inspired by or based on, in some manner or other, a previous work of art.

This is exactly what machine learning does - a novel recombination of what has been previously observed (ie., "reusing the music" - building upon that which is)

Whether this process is done by a human or not is irrelevant. ML, for the most part, does exactly what humans do.

There are currently very few capabilities which ML performs that humans do not - namely, blatant plagiarism.

People whining about their exact job being done by a machine is as old as time. Farmers, industrial revolution, the printing press (scribes whined), etc.

Nothing new here. New jobs will be created to replace the lost ones. We've arguably been in a "BS economy" for quite some time now.

[999900000999]

True, but nothing could ever be done so quickly and at such a large scale.

If an Italian band wants to make something inspired by Pink Floyd it's going to take real effort. Not a 30 second click to generate.

Eventually, and this is probably my biggest fear with AI, is the AI generated noise will crowd out real human innovation. I don't feel that AI generated art should have the same protections that human generated art has.

[dfxm12]

Personally, I don't think the amount of effort is what's important. If Roger Waters, or "his people", sniff out something fishy, they're going to sue and they're going to get something for their troubles. I think what's important is that that legal option is uniquely available to him because he's already rich. If someone rips me off, it doesn't matter if they're working with a trained ear or AI. There's nothing I can do about it cause I can't afford a lawyer.

There should be a better legal framework for us regular folk who don't have these resources, regardless of AI. A better framework probably includes what you're saying about AI generated things, but there's no reason for it to stop there.

[gedy]

> AI generated noise will crowd out real human innovation

But that’s the key point, “real human innovation” will still be there because by definition that’s not going to be in some model weights based on previous works. If it’s truly innovative, AI won’t be able to duplicate that.

[yowzadave]

If we outsource the job of art-making to bots, what will remain for humans to do? Given that there are so many jobs that involve real drudgery, it’s odd to me that we’re trying so hard to automate the fun/rewarding/creative parts, while leaving the difficult, repetitive, mind-numbing parts to humans.

[bee_rider]

I think this is probably because interesting jobs require handling information, which computers are good at, and boring ones require just shuffling stuff around in the physical world, which computers are still mediocre at.

But, designing a robot that can do tedious tasks is an intellectual job, just a really hard one. If we automate intellectual jobs to the point where, like, any random team of 10 people with engineering degrees can make a really good robot, I guess we should see the tedious jobs taken too.

[yowzadave]

I used to work as an architect. As far as I am aware, AI/automation has so far only made inroads in the earliest "concept" phases--image-making, concept massing, conceptual space-planning, etc. That is only the first, shortest, and most creative part of designing a building. After the concept design has finished, there are longer and more expensive phases of design development and construction drawing that still require huge amounts of coordination between the various design/construction professionals & stakeholders. At this phase, the work is much more practical: adding detail to the drawings reveals that, say, the size of the required mechanical units exceed the area that originally been allocated to them in the concept phase, and the architect will need to work with the structural engineer/mechanical engineer/client/etc. to modify the solution so that the building still works. It's this latter phase of drawing/coordinating/redrawing that has so far been very resistant to automation.

I would still call this part of an architect's job "intellectual work". My guess is that it hasn't been automated because it requires deep domain knowledge in multiple areas...knowledge which your typical AI software person doesn't have.

[P_I_Staker]

What kind of "art work" is being AI'd? Art itself can be repetitive and mind numbing. I'd like some artists to weigh in, if they exist here.

It seems like in many people's minds art is something that comes from within. An idea pops into a brain; you ascend to the clouds. There it is, clear as day: YOUR IDEA. All you have to do is collect it and take it down to earth to sell... of course I don't have to pay you! You didn't have to do any REAL work. You're basically running through the fields catching butterflies. Seems like a common misconception about art.

The reality of art to me seems like it can be just as drudgy as doing the dishes, spreadsheets, or digging ditches. It seems like this is where we're seeing AI used the most?

It's interesting to see famous songwriters weigh in, many of which are likely glorified producers. From what, I understand this has sort of already happened to musicians, with the invent of DAWs and other tooling, even earlier versions of AI.

Wondering if they're finally coming for the engineers and producers. Actually these celeb artists will probably already be fine. They'll just pay their people to use AI, instead of paying session musicians and producers.

[raven105x]

That's an economic problem to be solved just like the printing press and the industrial revolution. Like I said, we've been in a BS economy for a while now.

[rurp]

> To be fair, this is no different than what's done by a human artist.

Scale matters. It's like the difference between a cop sitting on the side of a road writing down the cars he sees on a pad of paper versus cameras plastered everywhere tracking every individual with ML software. You could say the computers are just doing the same thing as the human, but one is fairly benign while the other enables a totalitarian surveillance state.

[lbotos]

And to be clearer -- We have licensing mechanisms now in music (they aren't perfect) but when you remix/interpolate/sample the original artist gets their cut.

The "risks" of Machine learning with large models is you won't know who you are even sampling, which exposes you to unknown risk.

[mft_]

You're right, but I don't believe the previous poster was talking about sampling.

What happens with machine learning models is more similar to someone learning to play music: everyone learning some form of musical expression will by default learn music which has been written previously, and this can't but help to influence them to a degree.

Of course, there are systems to police this too, but high profile cases in recent years have shown how imperfect (or difficult?) this is in reality [0].

[0] https://www.youtube.com/watch?v=NcCKlsTgjeM

[raven105x]

This is not true. Google LLMs / multi-modals fully support attribution and as explainability goes up, so will attribution.

And it will be much more fair / accurate than the droves of human artists who intentionally omit attributions.

When a human artist composes a new song, they often have no idea where they got the compound basis for it either.

If anything, human attribution is vastly inferior to ML as-is, nevermind of what's to come.

[mcguire]

Two problems here:

1. If a human does that, they can be sued for doing so. If the reuse is egregious enough they will lose in court.

2. Industrial farming and industrial production have in fact been some of the more contentious economic projects. And the results have often been a race to the bottom in terms of quality.