In A History Of The World In 10 1/2 Chapters by Julian Barnes, there's a chapter about a man who gets into Heaven, the realm with infinite possibilities: no boundaries, no unfulfilled desires. The eternal cycle of indulgence. The protagonist can do anything he could imagine. Food, travelling, sex, meeting Hitler (kind of) so on and so forth, every day for the rest of his life that never ends. But "after a while, getting what you want all the time is very close to not getting what you want all the time." Heaven stops being fun anymore. It becomes normal. Even novelty must not be abundant otherwise we start looking for a different orthogonal or reversed kind of it. The protagonist feels this, too, and chooses to leave Heaven and die permanently. "I dreamt that I woke up. It’s the oldest dream of all, and I’ve just had it," he says.
So, why I wrote all that is I think this applies to literally everything. Even excessive novelty is prone to inflation, and any abundance is satanic in a way :D Anything in abundance quickly becomes “boring”, less stimulating, “not sparking joy” and often we start looking for something new / unique / scarce. Not exactly in the same way, but I think something similar might happen to images or whatever AI can generate.
Although I agree Dall-E 2 will disrupt commercial illustrations industry at least somehow, I think it’s quite different for Art and it’s different for mass-produced vs specially crafted, just like it is now. I think we as artists will find a way to create something that stands out stylistically or with some other qualities, something surprising and stimulating, or at least something that has a mark “made by human” and costs x2 just like “organic” products now.
So, pardon me a repetition, but I want to highlight one of the ground truths of human existence, "after a while, getting what you want all the time is very close to not getting what you want all the time."
P.S. your suggestion on writing a short fiction is very tempting, I can’t think of anything else now.
All the styles Dalle creates in are pre existing styles of art that have been developed by humans in the past - is AI going to come up with new styles that appeal to humans do you reckon or is stylistic innovation still a human centric domain?
That pic of the Andy Warhol astronaut looks like what I would think an AI would think is a Warhol-esque style but I wouldn't say it's very aesthetic - why is it centred so weirdly on the page for one thing. There are a lot of blobby looking artefacts where the reins would be. The astronaut's arms aren't defined and kind of join onto the mane in a mess. Warhol's screen prints are neat and sharp.
The artwork is impressive on the whole. I'm impressed with how it looks for AI generated art. But the pencil drawings of horses are almost all horrendous and have weird growths, if you look at them closely. Anyway, those are some of my thoughts, but I am a commercial illustrator, so I'm feeling quite judgy.
You have to understand that you can generate an infinite number of astronaut paintings in the style of your favorite artist, one of which is bound to be flawless to your eye. We’ll be drowning in an infinite sea of content.
DALL-E 2 is even more powerful than that: you can generate variations on that specific image, and you can also edit sections of the image. Don't like those reins? Edit that section until you do. Check out the paper.
This isn't meant to replace human illustrators 1 to 1. Its more like instead of one person getting paid to develop an original image for a week you have someone coming up with the captions and the ai churning out infinite variations, and then a few people whose job is to quickly discard ones that are duds and save the ok ones, and then a bunch of people whose job it is to fix all the little glitches to make it look less ai generated. Not paid much and have to get a certain number done a day. Quanity. Efficiency. That's the kind of job the machine wants. Certainly not Art art. Until it figures out how to automate away the last two steps too.
Real purpose of AI like this is probably less about illustration and more about attention chaff. Make ten thousand things that look just like [ the thing it predicts you are interested in ] at least at first glance so the real thing becomes lost among them. Like the stacks of pink flowers in Horton Hears a Who Unless you pay for it maybe. Primitive versions exist now- the Taboola chumbox ads that look like news articles that would be interesting to you if you were a total moron, Google-spam webpages that generate a fake info blog about[ your search term], hastily made knockoffs of every popular game on the app store- now just imagine this perfected, done to everything everywhere.
In A History Of The World In 10 1/2 Chapters by Julian Barnes, there's a chapter about a man who gets into Heaven, the realm with infinite possibilities: no boundaries, no unfulfilled desires. The eternal cycle of indulgence. The protagonist can do anything he could imagine. Food, travelling, sex, meeting Hitler (kind of) so on and so forth, every day for the rest of his life that never ends. But "after a while, getting what you want all the time is very close to not getting what you want all the time." Heaven stops being fun anymore. It becomes normal. Even novelty must not be abundant otherwise we start looking for a different orthogonal or reversed kind of it. The protagonist feels this, too, and chooses to leave Heaven and die permanently. "I dreamt that I woke up. It’s the oldest dream of all, and I’ve just had it," he says.
So, why I wrote all that is I think this applies to literally everything. Even excessive novelty is prone to inflation, and any abundance is satanic in a way :D Anything in abundance quickly becomes “boring”, less stimulating, “not sparking joy” and often we start looking for something new / unique / scarce. Not exactly in the same way, but I think something similar might happen to images or whatever AI can generate.
Although I agree Dall-E 2 will disrupt commercial illustrations industry at least somehow, I think it’s quite different for Art and it’s different for mass-produced vs specially crafted, just like it is now. I think we as artists will find a way to create something that stands out stylistically or with some other qualities, something surprising and stimulating, or at least something that has a mark “made by human” and costs x2 just like “organic” products now.
So, pardon me a repetition, but I want to highlight one of the ground truths of human existence, "after a while, getting what you want all the time is very close to not getting what you want all the time."
P.S. your suggestion on writing a short fiction is very tempting, I can’t think of anything else now.
Cheers
All the styles Dalle creates in are pre existing styles of art that have been developed by humans in the past - is AI going to come up with new styles that appeal to humans do you reckon or is stylistic innovation still a human centric domain?
That pic of the Andy Warhol astronaut looks like what I would think an AI would think is a Warhol-esque style but I wouldn't say it's very aesthetic - why is it centred so weirdly on the page for one thing. There are a lot of blobby looking artefacts where the reins would be. The astronaut's arms aren't defined and kind of join onto the mane in a mess. Warhol's screen prints are neat and sharp.
The artwork is impressive on the whole. I'm impressed with how it looks for AI generated art. But the pencil drawings of horses are almost all horrendous and have weird growths, if you look at them closely. Anyway, those are some of my thoughts, but I am a commercial illustrator, so I'm feeling quite judgy.
You have to understand that you can generate an infinite number of astronaut paintings in the style of your favorite artist, one of which is bound to be flawless to your eye. We’ll be drowning in an infinite sea of content.
DALL-E 2 is even more powerful than that: you can generate variations on that specific image, and you can also edit sections of the image. Don't like those reins? Edit that section until you do. Check out the paper.
This isn't meant to replace human illustrators 1 to 1. Its more like instead of one person getting paid to develop an original image for a week you have someone coming up with the captions and the ai churning out infinite variations, and then a few people whose job is to quickly discard ones that are duds and save the ok ones, and then a bunch of people whose job it is to fix all the little glitches to make it look less ai generated. Not paid much and have to get a certain number done a day. Quanity. Efficiency. That's the kind of job the machine wants. Certainly not Art art. Until it figures out how to automate away the last two steps too.
Real purpose of AI like this is probably less about illustration and more about attention chaff. Make ten thousand things that look just like [ the thing it predicts you are interested in ] at least at first glance so the real thing becomes lost among them. Like the stacks of pink flowers in Horton Hears a Who Unless you pay for it maybe. Primitive versions exist now- the Taboola chumbox ads that look like news articles that would be interesting to you if you were a total moron, Google-spam webpages that generate a fake info blog about[ your search term], hastily made knockoffs of every popular game on the app store- now just imagine this perfected, done to everything everywhere.