Here’s my reason for it. Let’s suppose that I have set a xylophone up outside near a rocky cliff face, and one day, some rocks fall loose from the top of the cliff and strike the xylophone in such a way as to coincidentally produce the melody of Bach’s Prelude in C Major from the Well-Tempered Clavier. Is this melody any less beautiful, less artistic, because it was not produced by a human? Does it really matter whether the xylophone event happened before or after Bach’s writing of the Prelude? If the xylophone event happened first, would we say Bach’s authoring of the melody was superfluous?
Consider this: there are 8 notes to a major scale, and so this means that there are only 32,768 possible 5-note sequences within one octave to make a melody out of (more if you count the timing of the notes, but the point remains). The possibility space of melodies is already implicitly formed by the medium. When Bach writes a 5 note melody, we say that he has created a melody - but we could just as well say that he discovered one of the pre-existing 32,768 melodies of 5 notes.
This paradigm is true in visual arts as well. We can start with a small example: imagine a community of pixel artists making black and white pixel art images on a canvas of 32x32 pixels. Or you could imagine them as weavers of rugs with up to 32 weaves in and out in both directions, if you’d rather a low-tech example. There are a HUGE number of possible ways to choose to color in these pixels even just black and white. But the number is still finite. Now let me ask you this. Have you ever made visual art before? If you have, you probably know how the blank canvas full of possibilities quickly narrows down to constraints as your composition comes along. “If my figure is posed like this, I can’t show both the elbow and wrist, unless I use a strange perspective…”, “if I give them black hair, it darkens the composition too much and doesn’t look as good, but maybe if I add more light it could work…” Etc. What is it that you’re doing as an artist? You’re narrowing down the possibilities, from the HUGE possibility space of the blank canvas, to narrower and narrower “acceptable” configurations according to the criteria of the goal you have in mind.
Now suppose instead that I was doing really constrained pixel art - black or white only on a 3x3 grid. In that case there are only 512 possible artworks to be made. In that case, we COULD lay out all 512 of them, and just pick the one we like best. But if we were not very smart people, maybe we couldn’t figure out this trick, and we’d have to use our artistry to explore the 512 possible canvases one by one. We can imagine an artist eventually choosing configuration #371 as their artwork. They probably won’t think of as though they’ve chosen configuration #371, they probably will think of it like “I have come up with this new arrangement of pixels on the 3x3 canvas” - but in reality all they did was discover a possibility that has already existed since the beginning of time. Either way, I hope you and I agree that this person’s pixel art, despite being small and likely pretty boring, is still ART. It’s a work of art, although maybe not a great one. Now if I have a computer do the same process - explore this latent possibility space according to some criteria, finally selecting one possible configuration - and let’s say the computer also selects #371. Are we going to say this is not art? But this would be paradoxical! It’s the same image the artist made! Anyone who is familiar with the notion of “the death of the author” will see this is quite the same sort of principle. And if the computer happened to select #371 before any human did, would we then accuse the human of having “copied” the computer? Clearly not. This line of thinking, to me, is a strong one to defend AI images as possibly being legitimate and original art.
As an artist, you cannot create a new possibility within the medium. You can only actualize a possibility that has always latently been implied by the constraints of that medium. This is why many musicians and artists often talk about “finding” a melody or “finding” a vision. They find it because they are searching. They are searching their own unique path through that massive possibility space. The possibility space is too large for us to just simply look at every possibility and pick the one we like best - so we have to explore it, choosing at every moment which direction is best to step towards next, based on what we’ve got so far, and what we think we’ve learned about the shape of this possibility landscape over our experiences as artists.
Here’s my reason for it. Let’s suppose that I have set a xylophone up outside near a rocky cliff face, and one day, some rocks fall loose from the top of the cliff and strike the xylophone in such a way as to coincidentally produce the melody of Bach’s Prelude in C Major from the Well-Tempered Clavier. Is this melody any less beautiful, less artistic, because it was not produced by a human? Does it really matter whether the xylophone event happened before or after Bach’s writing of the Prelude? If the xylophone event happened first, would we say Bach’s authoring of the melody was superfluous?
Consider this: there are 8 notes to a major scale, and so this means that there are only 32,768 possible 5-note sequences within one octave to make a melody out of (more if you count the timing of the notes, but the point remains). The possibility space of melodies is already implicitly formed by the medium. When Bach writes a 5 note melody, we say that he has created a melody - but we could just as well say that he discovered one of the pre-existing 32,768 melodies of 5 notes.
This paradigm is true in visual arts as well. We can start with a small example: imagine a community of pixel artists making black and white pixel art images on a canvas of 32x32 pixels. Or you could imagine them as weavers of rugs with up to 32 weaves in and out in both directions, if you’d rather a low-tech example. There are a HUGE number of possible ways to choose to color in these pixels even just black and white. But the number is still finite. Now let me ask you this. Have you ever made visual art before? If you have, you probably know how the blank canvas full of possibilities quickly narrows down to constraints as your composition comes along. “If my figure is posed like this, I can’t show both the elbow and wrist, unless I use a strange perspective…”, “if I give them black hair, it darkens the composition too much and doesn’t look as good, but maybe if I add more light it could work…” Etc. What is it that you’re doing as an artist? You’re narrowing down the possibilities, from the HUGE possibility space of the blank canvas, to narrower and narrower “acceptable” configurations according to the criteria of the goal you have in mind.
Now suppose instead that I was doing really constrained pixel art - black or white only on a 3x3 grid. In that case there are only 512 possible artworks to be made. In that case, we COULD lay out all 512 of them, and just pick the one we like best. But if we were not very smart people, maybe we couldn’t figure out this trick, and we’d have to use our artistry to explore the 512 possible canvases one by one. We can imagine an artist eventually choosing configuration #371 as their artwork. They probably won’t think of as though they’ve chosen configuration #371, they probably will think of it like “I have come up with this new arrangement of pixels on the 3x3 canvas” - but in reality all they did was discover a possibility that has already existed since the beginning of time. Either way, I hope you and I agree that this person’s pixel art, despite being small and likely pretty boring, is still ART. It’s a work of art, although maybe not a great one. Now if I have a computer do the same process - explore this latent possibility space according to some criteria, finally selecting one possible configuration - and let’s say the computer also selects #371. Are we going to say this is not art? But this would be paradoxical! It’s the same image the artist made! Anyone who is familiar with the notion of “the death of the author” will see this is quite the same sort of principle. And if the computer happened to select #371 before any human did, would we then accuse the human of having “copied” the computer? Clearly not. This line of thinking, to me, is a strong one to defend AI images as possibly being legitimate and original art.
As an artist, you cannot create a new possibility within the medium. You can only actualize a possibility that has always latently been implied by the constraints of that medium. This is why many musicians and artists often talk about “finding” a melody or “finding” a vision. They find it because they are searching. They are searching their own unique path through that massive possibility space. The possibility space is too large for us to just simply look at every possibility and pick the one we like best - so we have to explore it, choosing at every moment which direction is best to step towards next, based on what we’ve got so far, and what we think we’ve learned about the shape of this possibility landscape over our experiences as artists.