Has anybody used AI at Folkrnn.org to make a "song"?

[Ralph124C41]

I ask because I watched a recent Youtube video by Marcel Ardans (of Lessons with Marcel) who use a "recurrent neural network" (RNN) , a form of artificial intelligence, to create folk tunes. His video is interesting because after several attempts he made what I thought to be a pretty good tune, one I liked so much that I got the tab for it and also listened to a companion short video in which a fiddler played it.

I went to the Folkrnn.org and played around with the tool. I still have virtually no idea of what I am doing but I did get to make some interesting tunes I thought, although not good enough to warrant saving them.

Anyway, I thought it is an interesting device. I understand some people have used RNN or another form of AI to create some interesting songs.

[Cecil6243]

Originally Posted by Ralph124C41

I ask because I watched a recent Youtube video by Marcel Ardans (of Lessons with Marcel) who use a "recurrent neural network" (RNN) , a form of artificial intelligence, to create folk tunes. His video is interesting because after several attempts he made what I thought to be a pretty good tune, one I liked so much that I got the tab for it and also listened to a companion short video in which a fiddler played it.

I went to the Folkrnn.org and played around with the tool. I still have virtually no idea of what I am doing but I did get to make some interesting tunes I thought, although not good enough to warrant saving them.

Anyway, I thought it is an interesting device. I understand some people have used RNN or another form of AI to create some interesting songs.

There's something about taking the realm of human creativity and letting a machine do it that bugs me. And how would a machine have those deep feelings of lost love, yearning for love, a heartfelt ballad etc.? Just my opinion of course.

[bleedingfingers]

Originally Posted by Cecil6243

There's something about taking the realm of human creativity and letting a machine do it that bugs me. And how would a machine have those deep feelings of lost love, yearning for love, a heartfelt ballad etc.? Just my opinion of course.

In the fascinating “Beatles key change” discussion just down the hall, JonPR points out that:

“they simply copied anything and everything they could lay their hands on. Their contemporaries were fully occupied with rock'n'roll (maybe just a few old standards), but the Beatles dipped into every popular genre across the board (and Paul and George even learned a classical tune or two).

… Basically, if you take Elvis Presley, Lonnie Donegan, Jerry Lee Lewis, Little Richard, Carl Perkins, the Everlys, Eddie Cochran, Chuck Berry, Roy Orbison, Fats Domino, Buddy Holly, Peggy Lee, the Shadows, the Miracles, the Isley Brothers, Ray Charles, the Shirelles, George Formby (yes!), Hank Williams, various 1920s jazz standards, etc etc, throw it all into a blender, add a drop of Scouse wit, and what comes out is: the Beatles - inevitably.”

Conceptually, this is exactly what generative AI like folkrnn is doing. It’s learning from thousands of examples of what was done in the past, and is generating new combinations of those common phrases

folkrnn has only learned from the common folk songs, though, not broad swathes of the entire breadth of western music like the above description re The Beatles…. But it’s the same essential process being described. One was directly 4 people synthesizing all their influences and recombining, the other is humans teaching a computer to do the same process for them…. It’s just a tool

[Cecil6243]

Originally Posted by bleedingfingers

In the fascinating “Beatles key change” discussion just down the hall, JonPR points out that:

“they simply copied anything and everything they could lay their hands on. Their contemporaries were fully occupied with rock'n'roll (maybe just a few old standards), but the Beatles dipped into every popular genre across the board (and Paul and George even learned a classical tune or two).

… Basically, if you take Elvis Presley, Lonnie Donegan, Jerry Lee Lewis, Little Richard, Carl Perkins, the Everlys, Eddie Cochran, Chuck Berry, Roy Orbison, Fats Domino, Buddy Holly, Peggy Lee, the Shadows, the Miracles, the Isley Brothers, Ray Charles, the Shirelles, George Formby (yes!), Hank Williams, various 1920s jazz standards, etc etc, throw it all into a blender, add a drop of Scouse wit, and what comes out is: the Beatles - inevitably.”

Conceptually, this is exactly what generative AI like folkrnn is doing. It’s learning from thousands of examples of what was done in the past, and is generating new combinations of those common phrases

folkrnn has only learned from the common folk songs, though, not broad swathes of the entire breadth of western music like the above description re The Beatles…. But it’s the same essential process being described. One was directly 4 people synthesizing all their influences and recombining, the other is humans teaching a computer to do the same process for them…. It’s just a tool

Good points. One is the human brain storing musical and life experiences and the other is a computer software.

Will I go out a buy an album that is totally AI generated? Doubtful.

[Andyrondack]

Originally Posted by bleedingfingers

In the fascinating “Beatles key change” discussion just down the hall, JonPR points out that:

“they simply copied anything and everything they could lay their hands on. Their contemporaries were fully occupied with rock'n'roll (maybe just a few old standards), but the Beatles dipped into every popular genre across the board (and Paul and George even learned a classical tune or two).

… Basically, if you take Elvis Presley, Lonnie Donegan, Jerry Lee Lewis, Little Richard, Carl Perkins, the Everlys, Eddie Cochran, Chuck Berry, Roy Orbison, Fats Domino, Buddy Holly, Peggy Lee, the Shadows, the Miracles, the Isley Brothers, Ray Charles, the Shirelles, George Formby (yes!), Hank Williams, various 1920s jazz standards, etc etc, throw it all into a blender, add a drop of Scouse wit, and what comes out is: the Beatles - inevitably.”

Conceptually, this is exactly what generative AI like folkrnn is doing. It’s learning from thousands of examples of what was done in the past, and is generating new combinations of those common phrases

folkrnn has only learned from the common folk songs, though, not broad swathes of the entire breadth of western music like the above description re The Beatles…. But it’s the same essential process being described. One was directly 4 people synthesizing all their influences and recombining, the other is humans teaching a computer to do the same process for them…. It’s just a tool

When you take a pallet of colours and mix them together you get brown, as demonstrated by the thread on AI "art".

[Cecil6243]

Originally Posted by Andyrondack

When you take a pallet of colours and mix them together you get brown, as demonstrated by the thread on AI "art".

Actually I'm disappointed there hasn't been more posts and debate.

[crw]

After mulling it over a bit, I guess I feel like if AI writes a good song, then there is another good song in the world. And people who can and want to write songs will continue to do so, and lots of them will be good, too.

Still, I have a certain unease with the idea of AI becoming ever more proficient at songwriting. Perhaps it will become so good, that it may churn out tons of songs per second, each of which we humans would find wondrously lovely and moving. Even the next Lennons or McCartneys would be lost in the noise.

It seems so good and so bad at the same time.

[srick]

Originally Posted by Andyrondack

When you take a pallet of colours and mix them together you get brown, as demonstrated by the thread on AI "art".

Or, if you add many ingredients into a stew and let it simmer, you can create a magnificent and nuanced blend that is tastier than each of its ingredients.

[Andyrondack]

Yes as long as there's a good cook making value judgements regarding what ingredients work well together.

[FrankHS]

Originally Posted by crw

...
Still, I have a certain unease with the idea of AI becoming ever more proficient at songwriting. Perhaps it will become so good, that it may churn out tons of songs per second, each of which we humans would find wondrously lovely and moving. Even the next Lennons or McCartneys would be lost in the noise.

It seems so good and so bad at the same time.

With the inevitable conditioning that's on the way, we're going to get over all the moral or aesthetic qualms, in a way similar to how the phoniness of online porn has never held back its apparent "success." The few holdouts will be inconsequential. For example, I stopped paying any attention to any of the voice competition TV shows after I learned that every single one of them uses pitch correction software for every "contestant" when needed. Anyone miss me? Top tier songwriters should still find a market alongside AI songwriting, but lower tier songwriters may not. Pop is (obviously) already pretty formulaic. Give the AI writer various fake human names, and we'll never know the difference.