Machine learning is a tool amongst many. That being said, most good art requires more than a single tool, tools should be used with care. If you use enough AI that it becomes part of your artistic identity, it’s unlikely that your work will be impactful.
I’m still waiting for someone to make art that requires machine learning and is obviously creative by our standards, instead of using AI to recreate old art. I know it’s possible to use this tool in a way that’s revolutionary, but the users and developers seem to have little interest in pushing art beyond replacing the artists.
I want to see someone develop an original ML model with an original training set that can generate something impossible by any other method. I have a feeling this kind of art would barely reach the mainstream, but it would outlast the slop.
Don’t hate the tools, hate what capitalism turns them into and uses them for.
Reminds me of the old panic that photography would be the death of painters. It was shortly followed by an all time boom in art and creativity as painters tried new things and moved on from photorealism.
There’s still so much room left for human art and artists even in a post-AI world, as long as we keep rejecting the slop and supporting actual artists. Then maybe new art forms will emerge. Who knows!
Brian Eno, Terry Riley, and John Cage are names that come close to doing what you are describing. The idea of “generative” or “stochastic” or “aleotoric” music has been around for longer than this current AI boom has.
I also found this fascinating bit of music on wiki:
There are possibilities, but there are 99 lazy and uncreative people who just want to press the “make music now” button for every 1 person that wants to spend hours building and training their own models/sequences. (Those 99 have absolutely ruined the lofi/study beats on YouTube…)
Let’s say the artist, first creates all the input that is fed to the AI for training.
Let it be sounds, films, objects, drawings, literature. Everything has to be created by the artist exclusively.
This will be a model that only knows the artist’s work and will generate output based on the work by the same artist.
Now, let’s do that in a community. Everyone is free to share their models with others. Every art created from there would list all models used.
Maybe someday we will have something like this. But we will only have this, if someone actively works on it, based on the way AI needs input. Something we are still learning and will sure change. We have to think of the AI we have now, like the first steps of humans actually building a functioning flying object. We are now at the step of the first set of wings, that keep us for 1 minute in the air, before failing and falling. That’s a long way until the first passenger airplane takes off.
I have a feeling that we will have to come up with new definitions of copyright in the future.
Have you not seen the AI generated QR code embedded in an image ? I don’t think it can be done without AI, Don’t know if you would consider it art, but I do : for example here is the first one I got when googling it https://www.qrafted.ai/img/gallery/girl-3.jpeg
Like all AI things unfortunately the web is flooded with them now…
Embedding the features of one image into another to create an illusion is a task I’d consider AI for, IF the artist performing that task can be propelled by using the output as a base. If it takes far more manual correction by artist to the point that it takes longer to make a finished piece, or if the time spent enjoying the process is diminished, it’s no longer worth it.
AI in art should be about automating the tasks that require scale or repetition, like how 3D graphics took much of the mathematical work from artists, letting them focus on sculpting their forms precisely.
Time freed from automating one task should be spent by the artist on another task, such that the work is done faster AND is appealing in a clear and obvious way.
The most “creative” way I’ve seen this done so far is using separate prompts for different 2d image elements in still painting, which appears to take longer to make less consistent results.
It feels like prompters rely on the divided tastes of the internet to convince people that their art looks good to someone, just not the current viewer.
![[Rusty Creates] 'Artists'](https://lemmy.world/pictrs/image/ee6ccbcc-d001-4b28-a88f-d39c2b75f860.png){kind=link}
Machine learning is a tool amongst many. That being said, most good art requires more than a single tool, tools should be used with care. If you use enough AI that it becomes part of your artistic identity, it’s unlikely that your work will be impactful.
I’m still waiting for someone to make art that requires machine learning and is obviously creative by our standards, instead of using AI to recreate old art. I know it’s possible to use this tool in a way that’s revolutionary, but the users and developers seem to have little interest in pushing art beyond replacing the artists.
I want to see someone develop an original ML model with an original training set that can generate something impossible by any other method. I have a feeling this kind of art would barely reach the mainstream, but it would outlast the slop.
Good take.
Don’t hate the tools, hate what capitalism turns them into and uses them for.
Reminds me of the old panic that photography would be the death of painters. It was shortly followed by an all time boom in art and creativity as painters tried new things and moved on from photorealism.
There’s still so much room left for human art and artists even in a post-AI world, as long as we keep rejecting the slop and supporting actual artists. Then maybe new art forms will emerge. Who knows!
Brian Eno, Terry Riley, and John Cage are names that come close to doing what you are describing. The idea of “generative” or “stochastic” or “aleotoric” music has been around for longer than this current AI boom has.
I also found this fascinating bit of music on wiki:
There are possibilities, but there are 99 lazy and uncreative people who just want to press the “make music now” button for every 1 person that wants to spend hours building and training their own models/sequences. (Those 99 have absolutely ruined the lofi/study beats on YouTube…)
That Machine Learning model will learn… from what?
The training data could be the same as it is today, but maybe something novel could be produced by changing the objective function?
That would be possible, in an abstract way.
Let’s say the artist, first creates all the input that is fed to the AI for training.
Let it be sounds, films, objects, drawings, literature. Everything has to be created by the artist exclusively.
This will be a model that only knows the artist’s work and will generate output based on the work by the same artist.
Now, let’s do that in a community. Everyone is free to share their models with others. Every art created from there would list all models used.
Maybe someday we will have something like this. But we will only have this, if someone actively works on it, based on the way AI needs input. Something we are still learning and will sure change. We have to think of the AI we have now, like the first steps of humans actually building a functioning flying object. We are now at the step of the first set of wings, that keep us for 1 minute in the air, before failing and falling. That’s a long way until the first passenger airplane takes off.
I have a feeling that we will have to come up with new definitions of copyright in the future.
Have you not seen the AI generated QR code embedded in an image ? I don’t think it can be done without AI, Don’t know if you would consider it art, but I do : for example here is the first one I got when googling it https://www.qrafted.ai/img/gallery/girl-3.jpeg
Like all AI things unfortunately the web is flooded with them now…
This is a very cool concept, but has anyone actually gotten this to read as a qr code?
I’ve tried a bunch of apps without any luck.
Yeah. I have the same question.
Embedding the features of one image into another to create an illusion is a task I’d consider AI for, IF the artist performing that task can be propelled by using the output as a base. If it takes far more manual correction by artist to the point that it takes longer to make a finished piece, or if the time spent enjoying the process is diminished, it’s no longer worth it.
AI in art should be about automating the tasks that require scale or repetition, like how 3D graphics took much of the mathematical work from artists, letting them focus on sculpting their forms precisely.
Time freed from automating one task should be spent by the artist on another task, such that the work is done faster AND is appealing in a clear and obvious way.
The most “creative” way I’ve seen this done so far is using separate prompts for different 2d image elements in still painting, which appears to take longer to make less consistent results.
It feels like prompters rely on the divided tastes of the internet to convince people that their art looks good to someone, just not the current viewer.