At what point does using tools to make art go from being creative to not being creative?
I think the video creator makes a lot of incorrect assumptions about how people create things like images using generative AI. It seems like he assumes people put in a single prompt, get a single image, and away you go. This is akin to quickly drawing something like stick figures without using an erasure, and being satisfied with what is produced. Is it art? In this case it probably doesn't matter what the answer is, because you probably just want something to do something.
If you're trying to make visual art using generative AI, you usually partake in an iterative process, similar to the process of sketching something, erasing bits, and sketching it again, but slightly differently. With AI, you give it a prompt, generate a few options, and then you can do different things like change the prompt and generate entirely new images, or take the seed of an image that looks generally like what you want and tweak the prompt with that seed, or you can take the image that is close and do img2img manipulation.
I think the key point that I was trying to make was not that AI is just a matter of a single prompt, but that by injecting AI into the process, you lose more and more space to express your own personality because AI covers the subtle differences in human expression when art is done in a more manual way.
For example, if you get 10 different people to draw a bird (from a real-life bird, let's say the bird was perched in front of them), you'd get ten very distinct drawings. Now, let's saw you get those 10 people to draw a different subject of their choosing (each one can choose something different): chances are it would be quite easy to pair the second 10 drawings to the first ten based on who drew them.
That free expression is lost with AI: you LOSE the individuality and mask it, because AI is necessarily an averaging or erasing of that individuality through combining the various artistic essences together. That essence is precisely the essence of creativity.
> by injecting AI into the process, you lose more and more space to express your own personality because AI covers the subtle differences in human expression when art is done in a more manual way.
This is true of every tool used for art. Markers lead to people commonly using the same colors. Paint brushes lead to details appearing to be brushes on. The printing press lead to standardization of typography for textual art. Where is the line between good tools that enable human creativity by making it easier to express things, and bad tools that make it too easy to express things?
> [...] That free expression is lost with AI: you LOSE the individuality and mask it
Are you sure? Have you tested it? I don't agree with that expectation. Different promoting techniques lead to different output styles. Different individuals develop different promoting techniques. Thus, individual style is not lost.
If all of the people drew stick figure style drawings with minimal effort, it would probably be just as difficult as it would be to pair AI generated images with the person who generated it, assuming they put as much effort into generating the image as the people who made the stick figure. If all of the people put just as much effort into generating AI images they liked as the effort of people who made an effort to draw somethinf nice, it would probably be just as easy to pair the AI generated images with the person who made it, as it is to pair drawings with the person who drew it.
> This is true of every tool used for art. Markers lead to people commonly using the same colors. Paint brushes lead to details appearing to be brushes on. The printing press lead to standardization of typography for textual art. Where is the line between good tools that enable human creativity by making it easier to express things, and bad tools that make it too easy to express things?
AI has already crossed that line. Where's the line between consciousness and not when a human zygote grows? Perhaps the line is blurred like that.
> Where's the line between consciousness and not when a human zygote grows? Perhaps the line is blurred like that.
This line is not blurred at all among individuals who have formed an opinion on the matter. Some people thing the line is at birth. Some people think the line is at conception. Some people think the line is at 3 months. Some people think the line is when a heartbeat can be detected. The line is only blurred at scale.
So I'm asking you: where is the line between good tools that enable human creativity by making it easier to express things, and bad tools that make it too easy to express things? Because if you can't answer that, we can't actually have a discussion about this topic.
Also any response on the rest of the comment would be appreciated if you have one.
> So I'm asking you: where is the line between good tools that enable human creativity by making it easier to express things, and bad tools that make it too easy to express things? Because if you can't answer that, we can't actually have a discussion about this topic.
Well, I do think there's at least a sufficient requirement in crossing that line: it is when the tool starts to give input into your creation based on what you are drawing/doing.
For example, yes, the computer takes away the artistry and writing by hand but if I'm in a text editor, it won't suggest a particular story based on my additional ideas any more than a pen would or a charcoal on rock.
Another example: if I use a film camera versus digital, I basically see the same thing through the viewfinder before I even press the shutter.
Or paiting: whether I'm using my fingers or a brush, I don't see any way to proceed but through my imagination.
AI crosses that line because it is like a sophisticated paint by numbers: it constrains you by guiding you down a path through its preconceived ideas. It is like a "choose your own adventure" book, if you've ever read one. It's not the same as actually going on an adventure.
> If all of the people put just as much effort into generating AI images they liked as the effort of people who made an effort to draw somethinf nice, it would probably be just as easy to pair the AI generated images with the person who made it, as it is to pair drawings with the person who drew it.
I do understand your point here, it does make sense. I guess that would have to be tested. But it would have to be tested with the 10 people told to draw the SAME subject matter, not different subject matter. I think in this case an actual experiment would need to be done:
> If all of the people drew stick figure style drawings with minimal effort, it would probably be just as difficult as it would be to pair AI generated images with the person who generated it, assuming they put as much effort into generating the image as the people who made the stick figure.
Also, I rather disagree wtih this: it's surprising how much style is put into even fairly minimal stick figure drawings. But again, I think this would be something that would be good to experiment with.
Perhaps one could assign a quantitative measure to "creative potential of a tool" using some combination of time spent on the drawing and the entropy of the final product or something.. (just throwing ideas out there).
The art stops being creative right at the moment when it dawns on you that the other players in the CounterStrike match are bots the game is using as placeholders until real players join.
10 comments
[ 4.8 ms ] story [ 36.9 ms ] threadI think the video creator makes a lot of incorrect assumptions about how people create things like images using generative AI. It seems like he assumes people put in a single prompt, get a single image, and away you go. This is akin to quickly drawing something like stick figures without using an erasure, and being satisfied with what is produced. Is it art? In this case it probably doesn't matter what the answer is, because you probably just want something to do something.
If you're trying to make visual art using generative AI, you usually partake in an iterative process, similar to the process of sketching something, erasing bits, and sketching it again, but slightly differently. With AI, you give it a prompt, generate a few options, and then you can do different things like change the prompt and generate entirely new images, or take the seed of an image that looks generally like what you want and tweak the prompt with that seed, or you can take the image that is close and do img2img manipulation.
For example, if you get 10 different people to draw a bird (from a real-life bird, let's say the bird was perched in front of them), you'd get ten very distinct drawings. Now, let's saw you get those 10 people to draw a different subject of their choosing (each one can choose something different): chances are it would be quite easy to pair the second 10 drawings to the first ten based on who drew them.
That free expression is lost with AI: you LOSE the individuality and mask it, because AI is necessarily an averaging or erasing of that individuality through combining the various artistic essences together. That essence is precisely the essence of creativity.
This is true of every tool used for art. Markers lead to people commonly using the same colors. Paint brushes lead to details appearing to be brushes on. The printing press lead to standardization of typography for textual art. Where is the line between good tools that enable human creativity by making it easier to express things, and bad tools that make it too easy to express things?
> [...] That free expression is lost with AI: you LOSE the individuality and mask it
Are you sure? Have you tested it? I don't agree with that expectation. Different promoting techniques lead to different output styles. Different individuals develop different promoting techniques. Thus, individual style is not lost.
If all of the people drew stick figure style drawings with minimal effort, it would probably be just as difficult as it would be to pair AI generated images with the person who generated it, assuming they put as much effort into generating the image as the people who made the stick figure. If all of the people put just as much effort into generating AI images they liked as the effort of people who made an effort to draw somethinf nice, it would probably be just as easy to pair the AI generated images with the person who made it, as it is to pair drawings with the person who drew it.
AI has already crossed that line. Where's the line between consciousness and not when a human zygote grows? Perhaps the line is blurred like that.
This line is not blurred at all among individuals who have formed an opinion on the matter. Some people thing the line is at birth. Some people think the line is at conception. Some people think the line is at 3 months. Some people think the line is when a heartbeat can be detected. The line is only blurred at scale.
So I'm asking you: where is the line between good tools that enable human creativity by making it easier to express things, and bad tools that make it too easy to express things? Because if you can't answer that, we can't actually have a discussion about this topic.
Also any response on the rest of the comment would be appreciated if you have one.
Well, I do think there's at least a sufficient requirement in crossing that line: it is when the tool starts to give input into your creation based on what you are drawing/doing.
For example, yes, the computer takes away the artistry and writing by hand but if I'm in a text editor, it won't suggest a particular story based on my additional ideas any more than a pen would or a charcoal on rock.
Another example: if I use a film camera versus digital, I basically see the same thing through the viewfinder before I even press the shutter.
Or paiting: whether I'm using my fingers or a brush, I don't see any way to proceed but through my imagination.
AI crosses that line because it is like a sophisticated paint by numbers: it constrains you by guiding you down a path through its preconceived ideas. It is like a "choose your own adventure" book, if you've ever read one. It's not the same as actually going on an adventure.
I do understand your point here, it does make sense. I guess that would have to be tested. But it would have to be tested with the 10 people told to draw the SAME subject matter, not different subject matter. I think in this case an actual experiment would need to be done:
Step 1: 10 people "draw" a bird with AI
Step 2: 10 people draw soemthing else with AI
And then get other poeple to pair the drawings...
Also, I rather disagree wtih this: it's surprising how much style is put into even fairly minimal stick figure drawings. But again, I think this would be something that would be good to experiment with.
Perhaps one could assign a quantitative measure to "creative potential of a tool" using some combination of time spent on the drawing and the entropy of the final product or something.. (just throwing ideas out there).