ai music The Future of AI-Generated Music

Hey Guys,
I hope you're doing well! I have been thinking about AI in music for a while, and I hope it's not a problem to share my thoughts about this field. Here's my (unsolicited) opinion on AI-generated music.

Human artists cannot be (completely) replaced by current AI technologies because these technologies lack taste, consciousness, and individuality, and they are unable to produce anything truly new. Additionally, training an AI on 10,000 pieces of good music does not guarantee that it will produce good music itself.

However, AI-generated music can be envisioned in projects where quality is secondary and budgets are low. Using AI music can reduce costs, but even a mediocre or slightly bad human composer can often produce better music for minimal pay. Human composers are consciously capable of improving or modifying compositions based on instructions, which is a capability that AI is only limitedly capable of.

So, while AI-generated music can be suitable for projects with low budgets and where quality is not a top priority, it cannot fully replicate the unique creativity and quality of human-created music.
Of course, these statements concern today's AI technology with neural networks. Neural networks are just mathematical matrices with weights adjusted by error-finding and error-correcting methods.

The current AI technology used requires the algorithm that forms the weights in the matrix to receive the correct outcome and perform error correction in each iteration accordingly. However, in the case of music, the correct outcome is partly subjective and the result depends on the musical materials used for training. However, according to current laws, copyrighted materials cannot be used to train music neural networks.

It's an interesting case; I'm on Fiverr, where I was offered $25 by a music-generating website for a complete 3-4 minute symphonic piece. Who would create a full symphonic composition for $25? Certainly not Danny Elfman or Hans Zimmer.

IMO this would be better in Composer's Headquarters maybe?

Anyway, I believe that AI music and art are largely positive developments.

In the arts, automation has only ever increased the value of skilled craftsmen who can "do it for real". In the late '90s, autotune was heralded as the death of the singer, but people who were actually able to sing have only become more valued by society and the highest-paying positions in the industry. 

This kind of thing does not happen in blue-collar industries. Once the robot can do your job, it's over — because not so deep down, and despite a lot of the frankly borderline-communist praise that the working-class enjoy today, people do not really value menial work, but rather culture.

It is also the case that AI seeks to emulate high standards in art and music. It is not seeking to recreate some modern art nonsense or atonal gibberish but rather the expertise of John Williams and Karl Bryullov. 

Even if it reaches that level, it will not be able to offer the human input and working relationship required to achieve a unified artistic vision. 

We are now entering an age where your cellphone can produce a better artistic work or piece of music in seconds than people taping bananas to walls and call it "art" or the myriad zimmer clones and "it's all subjective" types, and these people will no longer be taken seriously by the masses or people with money to spend.

They will either have to start taking the craft seriously and "git gud" or go the way of the dodo.

1 hour ago, AngelCityOutlaw said:

IMO this would be better in Composer's Headquarters maybe?

Anyway, I believe that AI music and art are largely positive developments.

In the arts, automation has only ever increased the value of skilled craftsmen who can "do it for real". In the late '90s, autotune was heralded as the death of the singer, but people who were actually able to sing have only become more valued by society and the highest-paying positions in the industry. 

This kind of thing does not happen in blue-collar industries. Once the robot can do your job, it's over — because not so deep down, and despite a lot of the frankly borderline-communist praise that the working-class enjoy today, people do not really value menial work, but rather culture.

It is also the case that AI seeks to emulate high standards in art and music. It is not seeking to recreate some modern art nonsense or atonal gibberish but rather the expertise of John Williams and Karl Bryullov. 

Even if it reaches that level, it will not be able to offer the human input and working relationship required to achieve a unified artistic vision. 

We are now entering an age where your cellphone can produce a better artistic work or piece of music in seconds than people taping bananas to walls and call it "art" or the myriad zimmer clones and "it's all subjective" types, and these people will no longer be taken seriously by the masses or people with money to spend.

They will either have to start taking the craft seriously and "git gud" or go the way of the dodo.

 

Thanks for your opinion! That's a fact, that AI can't invent or figure out anything. It can generate things based on its training set. If John Williams or Karl Bryullov don't consent, the AI music-generator algorithms just can't use their stuff. If an AI algorithm creates something extraordinary that would be just a coincidence, and probably unrepeatable, but a human can invent something, and use it consciously. However, I'm a quiet skeptic about your standpoint, that AI will lead us a new age that talentless idiots won't be able to spread their shi*t. Maybe it will be more interesting because the stupidity of mankind is not replicable by an AI.

Moved this to Composers HQ so that there's more visibility.

While I do agree AI can be a good thing, it's just a tool. I've always said this...a 2 year old is smarter than AI. The reason? Because a 2 year old has actual intelligence, and can make decisions based on external stimuli. Whereas, AI is not intelligence, it's a misnomer.  I would challenge an AI to create a melody at the Tchaikovsky level. The AI may create music that sounds like Tchaikovsky, but just the harmonies perhaps. This is largely because harmony is more of the mathematical aspect of music, whereas melody or structure is the creative aspect of music. Anything to do with creativity, is something AI wouldn't handle that well. It may simulate creativity, but it will never actually execute something on its own.

On 5/11/2024 at 5:13 PM, olivercomposer said:

Hey Guys,
I hope you're doing well! I have been thinking about AI in music for a while, and I hope it's not a problem to share my thoughts about this field. Here's my (unsolicited) opinion on AI-generated music.

Human artists cannot be (completely) replaced by current AI technologies because these technologies lack taste, consciousness, and individuality, and they are unable to produce anything truly new. Additionally, training an AI on 10,000 pieces of good music does not guarantee that it will produce good music itself.

However, AI-generated music can be envisioned in projects where quality is secondary and budgets are low. Using AI music can reduce costs, but even a mediocre or slightly bad human composer can often produce better music for minimal pay. Human composers are consciously capable of improving or modifying compositions based on instructions, which is a capability that AI is only limitedly capable of.

So, while AI-generated music can be suitable for projects with low budgets and where quality is not a top priority, it cannot fully replicate the unique creativity and quality of human-created music.
Of course, these statements concern today's AI technology with neural networks. Neural networks are just mathematical matrices with weights adjusted by error-finding and error-correcting methods.

The current AI technology used requires the algorithm that forms the weights in the matrix to receive the correct outcome and perform error correction in each iteration accordingly. However, in the case of music, the correct outcome is partly subjective and the result depends on the musical materials used for training. However, according to current laws, copyrighted materials cannot be used to train music neural networks.

It's an interesting case; I'm on Fiverr, where I was offered $25 by a music-generating website for a complete 3-4 minute symphonic piece. Who would create a full symphonic composition for $25? Certainly not Danny Elfman or Hans Zimmer.

AI in art can be both useful and not so useful, because it can create some confusing moments. However, AI can definitely be useful in the IT sector. For example, creating an ai powered crm system https://ddi-dev.com/blog/programming/ai-in-crm-benefits-and-use-cases-of-ai-powered-crm-system/ can really advance your project and help you update it. It is quite useful and interesting.

AI music, while impressive, still lacks the nuanced creativity and emotional depth that human composers bring to their work. Current AI tools, despite their ability to generate music, can't truly replicate the individuality and taste of human artists. They often produce outputs based on patterns in training data, which might not always translate into high-quality or innovative compositions.

On 8/29/2024 at 5:05 AM, Colin_Huynh said:

Current AI tools, despite their ability to generate music, can't truly replicate the individuality and taste of human artists.

The thing about that though is that no one cares.

Listeners do not care, anyone who would hire composers does not care, etc.

One thing that can be given to the individuality camp is that AI does not seem to be able to, and may never be able to, do something like compose an entire album that has a consistent style. So it composes one piece of music that you like, it can't compose more that retain the same "fingerprint" that Beethoven or Metallica has throughout their work. People will probably find that disappointing.

But not disappointing enough to turn people away from AI.

Because all that people ultimately care about is whether or not it sounds good.

However, it could be that this is all a moot discussion as the sustainability of AI is coming more and more into question. These programs require large data centers to maintain, and they are pretty much all spending more money than they are making; it's mostly investors keeping them above water. At some point prices of the services are going to skyrocket because they will need even more hardware, electricity and such to support literally everyone using just a handful of services. Private individuals, big companies, everyone.

Then you must consider something like ChatGPT. ChatGPT is not replacing search engines because why pay to ask the robot when just about anything you want to know has already been written by humans and is freely available online?

The other issue is that AI cannot presently make changes in small details; it has to reinvent the entire image or whatever. If it were capable of doing small, specific tweaks, this will exponentially increase the demand for processing power. So far, despite the fears that it will destroy all the jobs, it has not proven to be a more profitable, efficient or economical alternative to people.

2 hours ago, AngelCityOutlaw said:

the sustainability of AI is coming more and more into question. These programs require large data centers to maintain, and they are pretty much all spending more money than they are making; it's mostly investors keeping them above water. At some point prices of the services are going to skyrocket because they will need even more hardware, electricity and such to support literally everyone using just a handful of services. Private individuals, big companies, everyone.

Yeah, I read some time ago that Open AI (the company that created Chat GPT) is going bankrupt

2 hours ago, AngelCityOutlaw said:

So far, despite the fears that it will destroy all the jobs, it has not proven to be a more profitable, efficient or economical alternative to people.

Last year, a Spanish soccer team “hired” an AI to decide what players they were going to buy and sell and for what price etc, and they ended up having one of their worse results in the last fifteen years. Let’s hope something similar happens with music