AI and the future of the 'Artist' vs the 'Creator' - TADay Talk Radio Podcast

[Al Maurice]

Whether we like it or not, music making simply wouldn't have been possible without technology.

Now we use computers with DAWs and notation editors to make music using VIs. And before that pen and paper, instruments and tape.

What's the difference here, we just have one more tool to play with.

We will still make music, just in a different way.

 

[Mike Fox]

Whether we like it or not, music making simply wouldn't have been possible without technology.

Now we use computers with DAWs and notation editors to make music using VIs. And before that pen and paper, instruments and tape.

What's the difference here, we just have one more tool to play with.

We will still make music, just in a different way.

I think it will depend on how involved AI is going to be.

Are we talking AI that still requires musical know-how and significant human input? Or will it be so automated that your Grandma can press a button and create a track?

But i do i think you’re correct in that AI will be another tool, because a lot of musicians will be using it as more of a minor assistant rather than a full blown slave that does everything for them. I’d like to think so, anyway.

 

[Mike Fox]

Well, let's the AI music spread the world, to the point that people get bored and come back to real organic music. Some years ago an AI player beat the best Chess players and then what? Nothing, chess players are still competing and nobody cares about Deep Blue and friends.

But it's a fact, composers will have to "compose" with this new trend for a while.

I could definitely see this being a very likely scenario. AI hits the music scene, people soak it up, the market gets saturated, but then people realize it’s not as rewarding as writing music without it, or playing a real instrument for that matter.

The sustainability of AI saturation in music is what I question most.

I do believe that composing jobs will be automated for the most part though, but maybe that will inspire musicians to become better at their craft.

 

[Noeticus]

The real problem is when AI lawyers and legislators make music created by humans illegal.

Eric Idle said that he is not worried about artificial intelligence, but that he is worried about artificial stupidity.

 

[Soundbed]

View attachment Max Headroom.mp4

 

[turnerofwheels]

Some related points of (hopeful) clarity rather than answers, since I did dabble with this already in school:

1. I'd distinguish between AI and other kinds of algorithmic (I include aleatoric) music, which has been around for centuries in various forms. Generating an ostinato, whether you are doing it by hand or by computer--you know the effect, how to achieve it, and you follow the process. It's no coincidence that ostinato-driven minimalism in the 1960s folds over with process music. You're already following predetermined algorithms whether you draw it out by hand or tell the computer to generate it.

It's also a common misconception to say algorithms write the music. When I work with algorithmic music, I am creating an environment or situation, composing with codes/rulesets, computer or not. I still have an idea of what I'm getting because I created the conditions for the output. I might change the process and experiment with the outcome, and learn what the relationship between my changes and the output is.

AI definitely changes this quite a lot.. even though I can train AI, and work with it, and maybe over time, once I intuit how the AI works, get a better idea of what to expect when I train it and learn it's particular quirks, so to speak

 

[turnerofwheels]

2. I was in a different music department than what folks are talking about here.. All of these topics are at the forefront of some music schools. Xenakis and people at ircam+grm created the concepts that led to granular synthesis and time stretching as one easy example, not to mention process music etc...

Back in the 2000s I was hanging around grad friends in Providence who were playing with neural networks to generate music, and training AI to respond to reappropriated power gloves. It was about crafting narrative (change) with AI. The AI becomes more of a collaborator that you train and work with and that is the present approach (as far as I can tell, I might be more out of the loop now).


3. Music education (not necessarily school of course) is also perceptual/sensory training. I can't tell you how many times I have heard people transcribe music without 7ths because they never paid much attention to jazz harmonies, or heard buskers who played Take 5 in 4/4, or how many newbs can't differentiate a hi hat from a snare or pull out harmonies, even if they are still affected by them. The only way to really adequately get past all of that is to get hands on. Also, if the AI can merely imitate but not craft an appropriate narrative over time, composition skills would still be required to fill in those gaps.

You'd still have to check your mixes too. AI mastering does the job cheaply, but it's got nothing on a real mastering engineer who crafts the music to fit the narrative flow of a track.. even though times are harder on mastering engineers now

So adding to your questions, one question could be partially how much society values any complexity, narrative, flow or ornateness in detail in art whatsoever when someone can pay $15 a month to AIMusicFriends.com to generate an epic score for their cat video. Naïve curation could become the new kind of low quality work, as well as shoddy narrative direction

 

[shadowsoflight]

Really interesting topic. On the one hand, I like the idea of being able to integrate various art forms as a "creator" without having to be an expert at all (or any?) of them. I could, for instance, one day create an epic space battle to go along with one of my tracks, even though I don't have time to do modeling/animating/effects etc. (Okay, okay, I realize that this specific example might be a ways off :D )

On the other hand, I think these sorts of trends are concerning - because society as a whole doesn't do a great job of valuing people, just economics. I believe that there would still be a place for real composers, but that it would be very niche. In the short term we will see changes to how composers work - by using AI to enhance workflows - but eventually it will reach a tipping point where we are just curating and tweaking AI outputs, which will itself reach a tipping point where the music is great right off the bat. But I strongly believe we will be in for a massive (hopefully not too traumatic) societal shift somewhere in there, since this problem will start to apply to almost every type of work.

This discussion reminds me of a video by CGP Grey on the state of automation:

 

[turnerofwheels]

Also:

This may face attempts at regulation, if it's even possible to contain. We're talking about technology that makes today's deepfakes and already bonkers information warfare look like Byzantine trebuchets by comparison. When you can generate anything the eyes and ears take in, it won't be isolated to music or Disney movies

Optimistic response: Maybe artists could carve a path through that slog

Pessimistic realist response: Welcome to the final stage of the simulation. All hail our not benign, corporate, hyperreality generating AI overlords

 

[chocobitz825]

Automation has replaced plenty of industries…in the end has anyone really sworn off McDonald in favor of homemade or gourmet food when convenience is the priority? Is anyone talking about going back to the days when cars where built by people?

the average person won’t know and won’t care if AI makes their music. They care about convenience and entertainment. Music is not that important.

 

[darkogav]

Automation has replaced plenty of industries…in the end has anyone really sworn off McDonald in favor of homemade or gourmet food when convenience is the priority? Is anyone talking about going back to the days when cars where built by people?

the average person won’t know and won’t care if AI makes their music. They care about convenience and entertainment. Music is not that important.

What is an "average" person?

 

[DimensionsTomorrow]

I’m a professional translator in the finance industry and we are facing similar trends. Machine output is still not quite good enough for professional Japanese translation requiring a very high level of accuracy, but is getting scarily good. I’m already being given these tools by my company to improve efficiency/output, but some day it will probably be a matter of quality control (you need someone bilingual to check the output), rather than doing it from scratch. In our case, the people that are being hit hardest seem to be freelancers as we need them less and less.

I’m guessing the music industry (TV/Film) might end up with a similar model of in-house experts driving the tools and ensuring professional quality and probably doing a lot of non-music production work on top of it (which is what we do now as well).

 

[chocobitz825]

What is an "average" person?

The average consumer of music. Someone who doesn’t consider themselves an artist, or fanatic of music. A person who participates in the social experience of music, but has no extended interest in it like composers, performers and fanatics.

 

[Al Maurice]

AI in science has been used extensively for numerous years, as a tool for predicting and checking out premises. I can see that in music it could prove handy for quickly trying out thematic ideas before committing.

If it's used as a means for automation, then that's a whole different matter and may be a cause for concern.

 

[Bluemount Score]

What is an "average" person?

A person that's not on VI-Control

 

[AudioLoco]

We are totally screwed ... ideally there should be a UN level agreement to ban AI from any form of art.
As a law. Martial law.

There is nothing good that could come from it to humanity, apart from non artist entities looking to make easy money.

I look with disgust to developers working on this, it's not like they are working on a heart surgery machine or something...

 

[Dewdman42]

There are significant ways in which AI could theoretically be used to help artists be more productive with the parts of the process which are not raw creativity but are rote and methodical, and time consuming to do, leaving the artist free to create.

But…. First there are very few people in this world that actually understand how and where to distinguish between the creative part of art production and the procedural parts that could streamlined. Instead we see a lot of early AI efforts to actually replace the creative part; which I view as suicide of humanity if it ever gets good enough to replace people entirely from a practical perspective of money.

Secondly, even if they could distinguish between creativity and procedural assistance in AI, in the case of music production, a lot of paid people make a living mainly being paid for their time to do all the boring procedural parts with only a few brief moments occasionally for the raw creative energy to come out. If those boring procedures are handled by AI, then there will be much less need for advanced education in music and 30 years of piano lessons, etc. There are scores of people out there possessing the raw creative energy to create wonderful music and other forms of art if they but had the procedural knowledge and skills to realize their creative ideas. So this would lead to an explosion of new artistic output but a lot of unemployed professionals that are no longer needed to realize human creativity.

However my observation is that most of the AI efforts seem to be trying to use AI to replace human creativity, which I think is a fatal error for humanity. So far it simply cannot compete with the human mind and possibly never will, but it might get good enough eventually that “good enough” original music will essentially become free and then we will never again have many humans spending time to develop true human creative musical ideas, because nobody will pay for it. It will just become a completely lost art form and music will just be an ever present commodity that is always there and nobody knows how or why

 

[Soundbed]

While some immediately think about how to "regulate" these tools, that impulse is foreign to me, sort of like regulating pencil and paper.

It's not a question of "when" AI will arrive at the doorstep of musicians. It's already here.

Mixing
The machine learning aspect of AI is already available in many of the mixing tools musicians use today including iZotope Ozone and Neutron.

I'm not sure which other mixing tools are using true AI (by definition) but there are plenty out there and again: they are already in use today by tens of thousands of musicians and producers.

The mix of a piece can greatly alter an audience's perception and emotional response.

Effects
Zynaptiq Adaptiverb is an example of a Reverb that uses AI. There are probably more. From an Attack Magazine article,

Zynaptiq calls Adaptiverb a “harmonic tracking resynthesis reverb” that uses a bionic sustain resynthesizer built using “an AI technique similar to that used to enable self-driving cars to avoid collisions”. It synthesizes a reverb tail using a network of hundreds of oscillators that learn to recreate just the pitched part of the incoming sound. It also has a ray tracing reverb module that again uses AI to simulate the paths sound waves take from a virtual sound source to a virtual listening position in a 3-D room model, except that it does 16,000 of these at the same time. We’d be lying if we said we knew exactly what this meant but one listen to the kinds of reverbs Adaptiverb is capable of proves the famous quotation by Issac Asimov that, “any sufficiently advanced technology is indistinguishable from magic”. from https://www.attackmagazine.com/reviews/the-best/the-best-ai-assist-plugins/

Choosing Sounds
Other tools help virtual drum programmers weed through their huge sample collections to select drum samples (which is arguably part of the creative process) like XLN Audio XO (https://www.xlnaudio.com/products/xo) and Algonaut Atlas (https://algonaut.audio/atlas-1-main/).

Apparently Jamahook Sound Assistant does this at an even more advanced level.

Choosing Pitches and Rhythms
Finally, tools are already on sale to help pick which notes to use in your compositions, too. Some of them even expect a bit of money in exchange for the right to use their "Composer Assistant" services. Examples include:

Orb Producer 3 – ORB COMPOSER

www.orb-composer.com

AIVA - The AI composing emotional soundtrack music

AIVA, the Artificial Intelligence music composer that creates original & personalized music for your projects.

aiva.ai

Flow Machines – AI assisted music production

AI assisted music production

www.flow-machines.com

From the Flow-Machines website:

Generally, the process of making a song includes composition, lyrics, arrangement, performance, mixing and mastering, but Flow Machines Professional specializes in composition. In recent years, in music production, a method called Co-Writing has been increasing, in which each creator, such as melody, track, arrangement, and songwriting, collaboratively composes by making use of their specialty. With Flow Machines Professional, we have created a new Co-Writing that fuses human creativity with AI music by selecting original melodies proposed by AI with the creator’s sense and promoting music production.