Yeah, I think it's mostly just for fun. Most things described as "Turing Tests" aren't much like what Turing had in mind, and don't answer a practical question either. Nobody would use a "Turing...
Yeah, I think it's mostly just for fun. Most things described as "Turing Tests" aren't much like what Turing had in mind, and don't answer a practical question either. Nobody would use a "Turing Test" to decide whether Waymos are safe enough to allow on the road. We wouldn't use a human driving test either. We often have higher standards for machines.
If anything, this just shows that with good enough human curation, AI art isn't that bad.
"With good enough curation, human art isn't that bad." Most human art is garbage as well. Nobody cares for the 300th crudely drawn Shadow the hedgehog posted to deviantArt. AI Art has shown me...
"With good enough curation, human art isn't that bad."
Most human art is garbage as well. Nobody cares for the 300th crudely drawn Shadow the hedgehog posted to deviantArt.
AI Art has shown me that the only form of art is that if of curation. Curation itself is an art and it is what is responsible for all other art. Art does not become famous on its own - it becomes famous due to the curators who decided it should be recognized. Curation can be a solo or group act and curation is meta in that people collectively choose the curators who can influence by their curation alone.
AI art is a matter of both solo curation and collective curation. People individually curate the art they generate and will post to share the art they think is good. Many people might post junk and those people are bad at curating. Once shared with other people (curators) good AI art that people collectively like will rise to the top and the rest will sink to the bottom until a curator plucks it from the bottom of the sea. It doesn't get treated any differently than human art in that regard. It will sink or float on its own merit and sometimes just pure luck that a collector found it in the pile of junk on the floor.
Curation also is meta in that people collectively curate what they think is worthy of curation and sometimes that's a popularity game in itself whether the banana you tape to a wall is art worthy of being curated or not.
Tildes has this oddly strong anti-AI slant as of late.
A traditional turning test does not include removing the batshit insane stuff you might convince an AI to say, so yeah I don't think this is really fair to say it's a "Turing test" It's especially...
A traditional turning test does not include removing the batshit insane stuff you might convince an AI to say, so yeah I don't think this is really fair to say it's a "Turing test"
It's especially awkward because how well you do will vary WILDLY on your personal knowledge of art, and knowing that AI can imitate some styles very easily (cute anime girl) due to various factors, while fail miserably at styles that are more esoteric or defined.
Consider just how insanely wide the gap on styles is for impressionism between the AI and the real art. Your average person might not even group those in the same genre of art, so it's a weird comparison to make to anyone but a real art critic/aficionado. Further the AI's output is easier for AI to do (scenery is generally easier for all sorts of reasons).
This is still interesting, but I don't think it shows nearly as much as people might imply (as with basically everything AI).
I'd like to echo this sentiment, it has been a long time ago but my art history classes I took way back did pay off here. There are differences in art styles, lack of meaning in the AI generated...
I'd like to echo this sentiment, it has been a long time ago but my art history classes I took way back did pay off here. There are differences in art styles, lack of meaning in the AI generated images also a few more subtle things like shadows being outright wrong.
I spend a lot more time looking at the anime girls and other more modern art styles I know are often AI generated. One exception is art styles that by themselves are quite busy, one thing AI generated images often have (without prompt refining) is an over-abundance in details. This made me fairly sure that the ancient gate picture was AI generated but didn't help me in other instances where the style itself is fairly busy in detail.
What also did not help is the fact that these images all have been scaled down. A lot of AI tells are in the details you can't quite see at the resolution of these images.
Having said that, it drives through the point that a lot of the use of AI generated imagery you see these days is outright lazy. Something I already knew on a certain level, having experimented with DALL-E and prompting. There is no real reason for most AI generated images to be that obviously AI generated. When I say lazy, I mean extremely lazy as well, getting a different output can be as simple as generate an image of X with Y, Don't add too many details, keep it simple potentially also adding a qualifier for the style you are looking for.
Here is an example of this at work with a woman staring into a sunset, here with a spaceship
I decided on both subjects as these seem to be popular subjects. If you know anything of the art styles you can still quite easily see that it isn't quite right. Certainly, the spaceship prompt would need a few more qualifiers for it to stand out less. At the same time, both initial prompts create something that give strong "AI Vibes" where the adjustments make it feel much less so.
Yes, it’s not a Turing Test, but people seem to use that term very loosely. I’m not sure how you would do a Turing Test for art? Maybe the judge sends a prompt, both the computer and the human...
Yes, it’s not a Turing Test, but people seem to use that term very loosely. I’m not sure how you would do a Turing Test for art? Maybe the judge sends a prompt, both the computer and the human artist have a few hours or a few days to make a picture, and then they judge the results?
It would be a more practical test for the AI side to be a human using generative AI, so they can generate as many images as they want within the time limit and take the best. That’s how these tools are actually used, after all.
But then, this doesn’t seem much like what Turing was trying to do. You could just as well call it a “John Henry” competition.
I just took this test myself and was surprised to only get 34/50. Yes, these examples were cherry picked, and many of them deliberately fool you, but still - as someone who's been playing with...
I just took this test myself and was surprised to only get 34/50. Yes, these examples were cherry picked, and many of them deliberately fool you, but still - as someone who's been playing with diffusion and image generation since the mid 2010s I was surprised to see how... not good, but high-fidelity the technology has become.
That being said however, the pictures here aren't really representative. They were generated, usually, by people that knew what they were doing. Were cognisant of "AI Tells" and tried to remove them. Knowing my way around the tools of the "trade" as well, I can absolutely guarantee some of these images got at least three passes in the inpainting mode to fix hands or details, and while that's valid as an artistic choice, it's not very representative of the AI generated "slop" that will clog up your google images and pinterests.
Side note: I personally believe if you go so far as to put in the actual effort to touch the piece up, either with Inpainting or Photoshop or literally printing and painting, you're sort of collaborating with the machine, making the piece AI-assisted rather than AI-generated. I think AI assistance is a perfectly valid technique and superficially no different than the clone stamp tool in Photoshop, so arguably to me those wouldn't qualify as "AI made" anyways.
What was even more interesting to me was that there were pictures in there that I could tell were generated, but still appealed to me on an artistic level. Notably, the Paris Scene one was genuinely quite enjoyable, and even knowing it was made with a machine learning model, I would have little trouble hanging it in my home and enjoying it. Maybe interpreting it like a human-made piece of art would defeat the purpose, sure, but just for decoration I think it's great.
What does this leave me with? Conflict, mainly. I know that this technology is neutral, as I've stated many times before, but the accessibility makes it easy for corporate ghouls to just bury us all in a goulash of extraneous fingers and nonsense greeble. I don't think AI art, whatever that means, is a threat to human creativity. That's ridiculous, and the same argument that game up when photography and later photomanipulation hit the scene. But I'd be lying if I said that we as a society have nothing to worry about.
I know a lot of people don't like Scott Alexander, but this seems like a fun post, with no politics any worse than a Sam Altman joke. And I thought the pictures in the contest were pretty...
I know a lot of people don't like Scott Alexander, but this seems like a fun post, with no politics any worse than a Sam Altman joke. And I thought the pictures in the contest were pretty interesting to look at.
From the blog post:
Last month, I challenged 11,000 people to classify fifty pictures as either human art or AI-generated images.
I originally planned five human and five AI pictures in each of four styles: Renaissance, 19th Century, Abstract/Modern, and Digital, for a total of forty. After receiving many exceptionally good submissions from local AI artists, I fudged a little and made it fifty. The final set included paintings by Domenichino, Gauguin, Basquiat, and others, plus a host of digital artists and AI hobbyists.
I participated in the contest and found it quite difficult, since images with obvious AI "tells" were removed. Perhaps someone with experience as an art critic would do better?
Personally, I found Scott's blog quite interesting (and I really-really love his Unsong book). I do not agree with some of his views/takes, but thats fine for me.
Personally, I found Scott's blog quite interesting (and I really-really love his Unsong book). I do not agree with some of his views/takes, but thats fine for me.
I don't know him personally, but AFAIC he's the best essayist I've read it years, and isn't because I agree with him all that often. He writes excellent think pieces that give you lots to chew on....
I don't know him personally, but AFAIC he's the best essayist I've read it years, and isn't because I agree with him all that often. He writes excellent think pieces that give you lots to chew on. I'm glad to see some of his work posted here.
I just took this, relying entirely on something that Niko from Corridor Digital pointed out in one of their videos: because image generation AI starts from Gaussion noise, the almost always end up...
I just took this, relying entirely on something that Niko from Corridor Digital pointed out in one of their videos: because image generation AI starts from Gaussion noise, the almost always end up with images that average out to 50% gray. That is, for every light region in the image, there must be a counterbalancing dark region.
I got 68% (34/50)! I may have been able to do better if I relied more on additional features, but that seems like a pretty decent indicator that, at least for now, that's not a bad rule of thumb.
Also just took this. The rot13 cipher answer key posted in the comments was missing some answers so I just didn’t include them when I graded myself (and I was too lazy to correlate to the actual...
Also just took this. The rot13 cipher answer key posted in the comments was missing some answers so I just didn’t include them when I graded myself (and I was too lazy to correlate to the actual images themselves since I was just on my tablet when I did this), but by my count I got about 74% correct. Honestly I kinda just went off vibes, like the article says particularly egregious examples of janky ai hands and such were missing, but issues like asymmetrical eyes, or subtle hand differences, or something I could best describe as “that odd ai image pattern” kinda gave it away for a few of them too (that goofy red and blue archway one was mainly the one I was thinking).
That was one where I was most surprised by his comment that he didn't see why it was AI. His artist-friend's commentary on unnecessary detail that makes no conceptual sense, and the lack of...
That was one where I was most surprised by his comment that he didn't see why it was AI. His artist-friend's commentary on unnecessary detail that makes no conceptual sense, and the lack of thematic coherence, were on point.
The thing with AI art is that, without sifting through unimaginable reams of training art, it's hard to tell how much of it is essentially plagiarism. Maybe the best pieces are basically just...
The thing with AI art is that, without sifting through unimaginable reams of training art, it's hard to tell how much of it is essentially plagiarism. Maybe the best pieces are basically just smoothed out collages of three or four existing pieces? I remember earlier on, there were many easily identifiable cases of this with generative text and art/video. With a much larger dataset, I'd imagine it would be quite difficult.
I was pretty bad at this, though I didn't bother to tally up my score. But I just want to comment that even though they say they took out art with misshapen hands, I saw several! Some of them were...
I was pretty bad at this, though I didn't bother to tally up my score. But I just want to comment that even though they say they took out art with misshapen hands, I saw several! Some of them were quite obvious if you were looking for them, so I'm curious how they were missed. I was definitely able to identify those as AI, although I was fooled by some of the other AI art because I thought the hands looked too natural. I also noticed a floating earring on one of the anime girl pictures.
Mostly I felt that if a picture was overly detailed with extra sharp lines, it was more likely to be AI. This definitely caught me out on the picture of the large ship, though I think I was right on some of the others. But for some of the other styles, like impressionist or modern art, I was left pretty clueless.
Not much of a Turing Test tbh, if it was curated and edited specifically to remove obvious tells on both sides.
Yeah, I think it's mostly just for fun. Most things described as "Turing Tests" aren't much like what Turing had in mind, and don't answer a practical question either. Nobody would use a "Turing Test" to decide whether Waymos are safe enough to allow on the road. We wouldn't use a human driving test either. We often have higher standards for machines.
If anything, this just shows that with good enough human curation, AI art isn't that bad.
"With good enough curation, human art isn't that bad."
Most human art is garbage as well. Nobody cares for the 300th crudely drawn Shadow the hedgehog posted to deviantArt.
AI Art has shown me that the only form of art is that if of curation. Curation itself is an art and it is what is responsible for all other art. Art does not become famous on its own - it becomes famous due to the curators who decided it should be recognized. Curation can be a solo or group act and curation is meta in that people collectively choose the curators who can influence by their curation alone.
AI art is a matter of both solo curation and collective curation. People individually curate the art they generate and will post to share the art they think is good. Many people might post junk and those people are bad at curating. Once shared with other people (curators) good AI art that people collectively like will rise to the top and the rest will sink to the bottom until a curator plucks it from the bottom of the sea. It doesn't get treated any differently than human art in that regard. It will sink or float on its own merit and sometimes just pure luck that a collector found it in the pile of junk on the floor.
Curation also is meta in that people collectively curate what they think is worthy of curation and sometimes that's a popularity game in itself whether the banana you tape to a wall is art worthy of being curated or not.
Tildes has this oddly strong anti-AI slant as of late.
Eh? I mean there are famous paintings like "Black Square" which might feel a bit unfair to include.
A traditional turning test does not include removing the batshit insane stuff you might convince an AI to say, so yeah I don't think this is really fair to say it's a "Turing test"
It's especially awkward because how well you do will vary WILDLY on your personal knowledge of art, and knowing that AI can imitate some styles very easily (cute anime girl) due to various factors, while fail miserably at styles that are more esoteric or defined.
Consider just how insanely wide the gap on styles is for impressionism between the AI and the real art. Your average person might not even group those in the same genre of art, so it's a weird comparison to make to anyone but a real art critic/aficionado. Further the AI's output is easier for AI to do (scenery is generally easier for all sorts of reasons).
This is still interesting, but I don't think it shows nearly as much as people might imply (as with basically everything AI).
I'd like to echo this sentiment, it has been a long time ago but my art history classes I took way back did pay off here. There are differences in art styles, lack of meaning in the AI generated images also a few more subtle things like shadows being outright wrong.
I spend a lot more time looking at the anime girls and other more modern art styles I know are often AI generated. One exception is art styles that by themselves are quite busy, one thing AI generated images often have (without prompt refining) is an over-abundance in details. This made me fairly sure that the ancient gate picture was AI generated but didn't help me in other instances where the style itself is fairly busy in detail.
What also did not help is the fact that these images all have been scaled down. A lot of AI tells are in the details you can't quite see at the resolution of these images.
Finally, some of the images that are human made also have been modified to make that less clear. This image does not look like the original
Having said that, it drives through the point that a lot of the use of AI generated imagery you see these days is outright lazy. Something I already knew on a certain level, having experimented with DALL-E and prompting. There is no real reason for most AI generated images to be that obviously AI generated. When I say lazy, I mean extremely lazy as well, getting a different output can be as simple as
generate an image of X with Y, Don't add too many details, keep it simplepotentially also adding a qualifier for the style you are looking for.Here is an example of this at work with a woman staring into a sunset, here with a spaceship
I decided on both subjects as these seem to be popular subjects. If you know anything of the art styles you can still quite easily see that it isn't quite right. Certainly, the spaceship prompt would need a few more qualifiers for it to stand out less. At the same time, both initial prompts create something that give strong "AI Vibes" where the adjustments make it feel much less so.
Yes, it’s not a Turing Test, but people seem to use that term very loosely. I’m not sure how you would do a Turing Test for art? Maybe the judge sends a prompt, both the computer and the human artist have a few hours or a few days to make a picture, and then they judge the results?
It would be a more practical test for the AI side to be a human using generative AI, so they can generate as many images as they want within the time limit and take the best. That’s how these tools are actually used, after all.
But then, this doesn’t seem much like what Turing was trying to do. You could just as well call it a “John Henry” competition.
I just took this test myself and was surprised to only get 34/50. Yes, these examples were cherry picked, and many of them deliberately fool you, but still - as someone who's been playing with diffusion and image generation since the mid 2010s I was surprised to see how... not good, but high-fidelity the technology has become.
That being said however, the pictures here aren't really representative. They were generated, usually, by people that knew what they were doing. Were cognisant of "AI Tells" and tried to remove them. Knowing my way around the tools of the "trade" as well, I can absolutely guarantee some of these images got at least three passes in the inpainting mode to fix hands or details, and while that's valid as an artistic choice, it's not very representative of the AI generated "slop" that will clog up your google images and pinterests.
Side note: I personally believe if you go so far as to put in the actual effort to touch the piece up, either with Inpainting or Photoshop or literally printing and painting, you're sort of collaborating with the machine, making the piece AI-assisted rather than AI-generated. I think AI assistance is a perfectly valid technique and superficially no different than the clone stamp tool in Photoshop, so arguably to me those wouldn't qualify as "AI made" anyways.
What was even more interesting to me was that there were pictures in there that I could tell were generated, but still appealed to me on an artistic level. Notably, the Paris Scene one was genuinely quite enjoyable, and even knowing it was made with a machine learning model, I would have little trouble hanging it in my home and enjoying it. Maybe interpreting it like a human-made piece of art would defeat the purpose, sure, but just for decoration I think it's great.
What does this leave me with? Conflict, mainly. I know that this technology is neutral, as I've stated many times before, but the accessibility makes it easy for corporate ghouls to just bury us all in a goulash of extraneous fingers and nonsense greeble. I don't think AI art, whatever that means, is a threat to human creativity. That's ridiculous, and the same argument that game up when photography and later photomanipulation hit the scene. But I'd be lying if I said that we as a society have nothing to worry about.
I know a lot of people don't like Scott Alexander, but this seems like a fun post, with no politics any worse than a Sam Altman joke. And I thought the pictures in the contest were pretty interesting to look at.
From the blog post:
I participated in the contest and found it quite difficult, since images with obvious AI "tells" were removed. Perhaps someone with experience as an art critic would do better?
Personally, I found Scott's blog quite interesting (and I really-really love his Unsong book). I do not agree with some of his views/takes, but thats fine for me.
I don't know him personally, but AFAIC he's the best essayist I've read it years, and isn't because I agree with him all that often. He writes excellent think pieces that give you lots to chew on. I'm glad to see some of his work posted here.
I just took this, relying entirely on something that Niko from Corridor Digital pointed out in one of their videos: because image generation AI starts from Gaussion noise, the almost always end up with images that average out to 50% gray. That is, for every light region in the image, there must be a counterbalancing dark region.
I got 68% (34/50)! I may have been able to do better if I relied more on additional features, but that seems like a pretty decent indicator that, at least for now, that's not a bad rule of thumb.
Also just took this. The rot13 cipher answer key posted in the comments was missing some answers so I just didn’t include them when I graded myself (and I was too lazy to correlate to the actual images themselves since I was just on my tablet when I did this), but by my count I got about 74% correct. Honestly I kinda just went off vibes, like the article says particularly egregious examples of janky ai hands and such were missing, but issues like asymmetrical eyes, or subtle hand differences, or something I could best describe as “that odd ai image pattern” kinda gave it away for a few of them too (that goofy red and blue archway one was mainly the one I was thinking).
That was one where I was most surprised by his comment that he didn't see why it was AI. His artist-friend's commentary on unnecessary detail that makes no conceptual sense, and the lack of thematic coherence, were on point.
The thing with AI art is that, without sifting through unimaginable reams of training art, it's hard to tell how much of it is essentially plagiarism. Maybe the best pieces are basically just smoothed out collages of three or four existing pieces? I remember earlier on, there were many easily identifiable cases of this with generative text and art/video. With a much larger dataset, I'd imagine it would be quite difficult.
I was pretty bad at this, though I didn't bother to tally up my score. But I just want to comment that even though they say they took out art with misshapen hands, I saw several! Some of them were quite obvious if you were looking for them, so I'm curious how they were missed. I was definitely able to identify those as AI, although I was fooled by some of the other AI art because I thought the hands looked too natural. I also noticed a floating earring on one of the anime girl pictures.
Mostly I felt that if a picture was overly detailed with extra sharp lines, it was more likely to be AI. This definitely caught me out on the picture of the large ship, though I think I was right on some of the others. But for some of the other styles, like impressionist or modern art, I was left pretty clueless.