It's more like having a contest that issues a prize to the person that hired the best artist to create a work. I guess it's as valid as any other arbitrary competition; we issue prizes to the owner of the fastest horse in a race regardless of who rides it.
Just look at the most left hand, it is so crooked.
But at least 2 of 3 hands are correct which is more than lucky based on my experience with stable diffusion. Usually the result is something like this image hosted on twitter: https://news.ycombinator.com/item?id=32875215
Just to clarify; I don't think AI-generated images have anything to do in a traditional photography contest. Prompt-creators can create their own festival.
I think it has some conceptual value - if many artists are saying that the ML models can't create real art and only get a mediocre rehash of existing ideas (which by definition can't be better than those ideas), then having the same artists judge whether some piece ML art actually is better than anything the traditional artists submitted kind of brings the theoretical argument down to reality.
Sure, they should be treated as separate categories of art, just as we don't include photographs in painting contests, however, that was solved by the artist promptly refusing the prize and disclosing the origin; but until this automatically dismissive attitude persists, triggering this discussion is a good thing - and I'd rather condemn the actions of the organizers, attempting to silently erase this event and pretending it didn't happen. Their jury did decide that at least in some cases an AI generated image can be better art than a "manually" taken photo, and they need to own that decision.
There's nothing wrong with just automatically dismissing AI created art in the context of a contest. We don't allow horses to enter marathons, because even though they're fast it's not for them. We don't even allow people to use performance enhancing drugs, which might be a closer comparison.
Should we "condemn" the judges of chess tournaments for not allowing live coaching and chess AIs?
My argument for condemnation was not for disallowing participation but about hiding/denying the fact that they did judge that artwork to be really good.
If the chess community would assert that computers are worse than humans as chess and couldn't ever match human play, and when someone won a chess tournament with forbidden computer assistance the judges of that tournament would instead of saying "yep, the win was due to a forbidden advantage, so we disqualified that participant" rather hide the fact and go "no, that didn't happen, obviously human reasoning won all the games", then we should absolutely condemn them for that hypocrisy.
"Just as photography replaced painting in the reproduction of reality, AI will replace photography. Don’t be afraid of the future. It will just be more obvious that our mind always created the world that makes it suffer."
ai and photography do not have the same purposes, so its unlikely AI will replace photography. Photography is used to record the world as it is, AI art creates depictions of what its asked for that have no relation to reality.
Rather what people were using photography for in the past may be replaced by AI. Before people went to stock photo sites for cheap images to put on their website or whatever, and tomorrow they'll find out it can be cheaper on midjourney or stable diffusion and it will do that.
But AI won't replace photography for what photography does, rather it will replace photography as the cheapest satisfactory image to use on your website that doesn't need a depiction of reality.
> Photography is used to record the world as it is
That's one use of photography. OP is about an art contest, not a photojournalism contest.
But as far as "reproduction of reality", obviously AI productions unreality won't replace photography. It might replace reproduction of reality, though.
I read it differently: the artist is saying (per my reading) that we humans have been making 'expressions of reality' first with paint and then with photographs. And now we will be using "AI", and what is good about this is how this latest mechanism of expression is laying bare the actual cause of suffering in human existence: our own minds and imagination.
photographs didn't destroy paint, because they don't do the same thing. People trying to represent the world as it is use photography (for the most part) and people wanting to express something about the world use paint.
AI is even less than that - people who want a computer to make their artistic decisions for them have a neural net author an image or other work for them. generally the humans having no more artistic input than someone who commissions a piece from an artist by explaining what they want.
A person painting makes a trillion times more artistic decision making than someone 'painting' with a neural net. They aren't the author of the work, and can't enter it into art competitions on their own behalf (unless the rules are stupid), and can't register it for copyright as there is no human authorship.
It is the complete dissolution of human art making - its humans deciding to have a non human thing - that cannot feel emotions or understand the emotions in an image - make 'art' for them. Most because they are incapable of doing it themselves so having something is better than having nothing, some because they do art primarily for money and the neural net works faster than them and they haven't figured out the bit about no copyright yet, or will misrepresent it to illegally get a copyright on an image they didn't author.
I'm not necessarily in disagreement with the general thrust of your comment.
That last point about feelings and emotions however is not clear cut, imo. Typically the process is itertive, so it's a loop and there is a feeling agent in that loop. Also if you yourself make art, you must have noticed how sometimes even the simplest of tools (say a brush or nib) create ~accidental features of the work that one finds entirely pleasing. This happens a lot to me with ink and brush :) The general input is distincly my line, it is my hand w/o a question, but I have a few works which feature engaging aspects that were partly accidental, and part of the delight is precisely that: me, the supposed author, is surprised.
>Typically the process is itertive, so it's a loop and there is a feeling agent in that loop.
Is someone who continually commissions art until they arrive at a painting they like making artistic decisions in the work? I don't think so.
>you must have noticed how sometimes even the simplest of tools (say a brush or nib) create ~accidental features of the work that one finds entirely pleasing. This happens a lot to me with ink and brush :) The general input is distincly my line, it is my hand w/o a question, but I have a few works which feature engaging aspects that were partly accidental, and part of the delight is precisely that: me, the supposed author, is surprised.
That is unquestionably something you authored, even the mistakes, even the mistakes you liked. But at no point of commissioning artwork are you making artistic decisions.
The 'commissioning' notion is how you wish to reduce the equation. I on the other hand consider AI as tool a reasonable proposition for the reasons noted. And from a purely practical point of view, getting worked up over this is tilting at windmills. Change is the constant. And there will be undoubtedly 'artist commissioners' that will outshine others, using the same 'artist'..
The commissioning notion was brought up by the copyright board in their march advisement on the eligibility of neural network produced works under copyright - which are not eligible.
It's an accurate description of whats going on. Commissioning paintings until you get the one you want isn't artistic expression, or artistic decision making, and the human coming up with prompts will be replaced by a machine learning program trained on prompts, as they bring the least value to the process and are the most easily replaced.
> And from a purely practical point of view, getting worked up over this is tilting at windmills.
I'm not worked up at all. I'm explaining whats going on. Is anyone who thinks differently from you getting worked up? Are you getting worked up for imagining emotions I'm not experiencing? In the future just argue ideas instead of imagining emotions across internet mind reading.
>Change is the constant. And there will be undoubtedly 'artist commissioners' that will outshine others, using the same 'artist'.
It will all approach zero value, so it doesn't matter. It's inauthentic, uncopyrightable stuff based on the misappropriation of the work of actual artists. The expert 'commissioners' will also be replaced by machine learning faster than anyone else.
Sorry, this image should not have won any prize. The hands are terrible, there are odd things around the back person's nose, and the highlights in the eyes are wrong, there is a reddish blob on the front person's head, and these are especially obvious when you look at it blown up a little. Whoever judged this contest is pretty clueless about the current quality of AI art generation and was easily fooled. I make digital art, not photography, but these jumped out at me.
The subject matter is probably what "won" the contest, it's pretty well composed.
AI images are not photos, obviously. We've conflated these words for a long time because the only way to create an image was by digitizing light. These AI images have more in common with a collage than a photo.
I don’t think a collage is the right way to describe it, because it implies that it’s literally parts of the training data copied together which a lot of people think but it’s a lot more indirect than that, but I agree that it’s not a photo because a photo is of a real thing
It’s more like the training data is used to score the generated images against to guide it into producing something that looks similar to a real image
I've told the story a few times on HN about the "cheating" in my 2007 Digital Imaging college photography contest for the Digital Imaging program[1]. In our instance, we're a totally (brand new) digital program, and someone combined some framed and created a scene that didn't exist, they shot the scenes separately and combined them. Our prof told folks to f'off when they complained as it was well within the lines. I agreed. That said...
This, seems outside the lines. Entering a photography contest with work that isn't based in photography, that you didn't shoot, isn't in the spirit of the contest. A photograph is an image created by light falling on a photosensitive surface - Edit all you want, hell edit till it's not even the same thing anymore, but if you're entering a photography content, at least trigger a shutter. The Sony photography awards are very clearly based around photography as the fundamental starting point. There are lots of pure digital art contests to enter, so enter them.
Agreed. For something like a photography contest, I am surprised they didn't have explicitly defined rules around it, like requiring contestants to provide the original RAW image (as a proof of work), having an explicit list of what is allowed to be tweaked (and nothing else would be allowed) like brightness/color grading/etc.
A lot of this predates generative AI. I know someone who used to be really into photography competitions and the like and he basically got discouraged because the photos he really worked on to get perfect right out of the camera were suddenly starting to lose to digital photos that had been enhanced in Photoshop in various ways.
Photos used for editorial purposes in newspapers and magazines at least historically had pretty strict rules about manipulation generally--including the removal of distracting elements that had nothing really to do with the core "truth" of the photo. I suppose the thinking was that manipulation can quickly become a slippery slope.
You're basically putting photos in front of people who've seen it all. They're not impressed by technical perfection or what gets 90+ on 500px.
edit: It's more likely your friend was losing out to people who could take more photos and be more experimental due to not being limited by the the cost of film or delay of development. Film can't touch digital's feedback loop.
Probably true of a lot of things. Food, theatre, art generally... Well-executed versions of something they've seen a million times before just isn't all that interesting.
I've entered photography contests before, and the ones I entered requested the raw file from the camera as well as your final .jpg/.png as a way of proving it was you at that took the image.
In regards to research I'm struggling to imagine useful applications of being able to create such raw files. Also, as it's a general term for a file containing raw sensor data, it would be different across sensors.
In any case, perhaps some kind of hardware-signed cryptography scheme on the files from the camera could be used in lieu of this development in the future.
I'll admit those are stretches. Anyway, in addition to a digital signing scheme, these competitions are gonna need more editing restrictions if they want to exclude the use of diffusion models.
The most obvious application would be faking criminal exhibits. Fake photos prrofing something
RAW files generated "backwards" do already exist. For this, police and forensics have been using special cameras with hardware signatures for raw files for around 15 years, but Canon's and Nikon's system was broken by Elcomsoft like 12 years ago. Sony is trying something similar in their A7 IV, but I don't know how serious it is, and I think it's not as oriented to forensics as Canon and Nikon was.
The article goes on to state that the Sony competition refused to acknowledge the withdrawal.
The author traveled to the ceremony at his own expense, found that he was not allocated time to speak, then went onstage unannounced to explain his refusal, at which point all physical evidence of his work was removed from the facility, again without comment.
I agree that Sony could have saved itself trouble by requiring the raw camera footage with verifiable exif data.
What is verifiable EXIF data? The submission was simulating a 100 year old photo. There could never be verifiable EXIF data for that. Sony didn't care; they liked the creativity.
Sony problem was in not catching someone who wanted to embarrass them. There's not a good way to detect that in advance.
Almost sounds like the collective decision making process of the awards organization were in denial that the industry at large is changing.
Feels dystopian that they would just, scrub the winning work from the website and physical location without comment, pretending as if it didn't happen.
I feel like the guy behaved pretty indecently, and to reward him with more attention would encourage others to behave similarly in the future, and detract from the other winners who participated in the contest in good faith.
I feel like he didn't behave indecently at all, and thus everything you said which follows, is not, in fact, the case
the fact that he was able to enter at all with a non-photo is proof that Sony is in the wrong, and despite their coverup, is not, in fact, prepared to deal with "black-hat artists" in this space
when an inadequacy makes itself apparent, the decent thing to do is acknowledge it, thank anyone who helped you see it, and try to make it right – Sony did none of these things, and thus are the "indecent" ones here
What sort of editing did you have in mind? If you are using SD to add things to the photo for example, you would see the difference between the raw file and the .jpg you submitted?
It’s already the case to add textures and patterns. Like sharpening images of the moon and its craters with AI. It’s hard to draw the line. Some phone does these AI work live now!
- Changing the style of the whole photo to (for instance) make it look old and analog.
- "Enhancing" it with HDR, depth of field, focus everywhere or something like that.
- Altering exposure beyond what you could get out of a underexposed/overexposed RAW file.
- Stylizing the human subjects to make them look more attractive, more grimey or whatever
All these things could arguably be done in photoshop, but its just very labor intensive and more hit-or-miss. Hence the contest rules would have to exclude most photoshop work if they want to exclude diffusion.
Even in traditional film photography, you can easily get "untrue" pictures. Film is sensitive to UV and near infrared, so filters can create very unreal imagery. And multiple exposure can do all sorts of tricks. The averaging of many faces together, creating a sort of uncanny valley result, was first done in the analog, stacking aligned photos with multiple exposure.
This is generally still considered photography. Art photography, if it's any good.
One could say the key element is the artist combining actual scenes that were exposed to film or a sensor. Perhaps massaged a bit in the dark room or digitally. But the core material is exposed images from actual scenes. It's photography while that element remains dominant. But then again ... isn't that sort of what these generative AIs also do? It's all originally derived from actual photographs.
the thing in pictures should happened in real world. I sure this is the difference between pictures and painting.
AI is famous in generate unreal pictures.
It is mind-boggling how the "trends" and the non-stop race for validation make all lose their minds.
How on earth a photo contest cannot see the false anatomy of the hands and person behind? Did someone in this concourse is actual photographer?
Maybe not.
So, on AI generators. I am an artist, I have painted and photographed all my life. The average prompter has no chance to win against a professional. I know which artist is presented in the data set, how to use the light and composition.
So basically speaking, this is the new Photoshop. The interesting question is what will hold long term more value in the future? The generated content or analogue product.
Can you describe what is wrong with the person behind? I can clearly see the hands issue and even I can see a strange hair on another person near her left ear. But the person behind seems correct to me.
Look at the dark eye, an artifact on nose, the lighting on the cheek behind shoulder. Fingers of her hand have abnormal thickness, index finger has very long nail, meaning they are not equal in length.
Thank you, now I see it all. This artifacts use to rhyme/mimicre with old-photo style and big artifacts on top-left quarter of image so they are not easy to see but now I see a usual stable diffusion style with noticeable disrespect to 3D objects everywhere.
> It is mind-boggling how the "trends" and the non-stop race for validation make all lose their minds. How on earth a photo contest cannot see the false anatomy of the hands and person behind? Did someone in this concourse is actual photographer? Maybe not.
I think you're trying very hard to poke holes at a very run-of-the-mill process. It's easy to pick apart stuff post-facto when you have the benefit of context and hindsight, and you are certain of what you're looking at.
If instead you're judging artwork, it would be unthinkable to reason "this work is pretty good but the model's fingers might look a bit off, and that is a good enough reason to reject the prize".
Throwing blanket accusations of incompetence at the jury also says more about yourself than the contest.
The truth of the matter is that AI -generated images are becoming quite impressive and can be a challenge to tell apart. This is not a problem with a jury in a context. We had deep fakes in the recent past, and they are improving rapidly.
It's easy to try to elevate yourself by putting down others, but while you're there making claims like it's obvious that the streaming audio of Biden and Obama and Trump and Hillary hanging out playing Minecraft is bullshit, multiple high-ranking officials already fell for calls from people using similar tech to impersonate world leaders.
Trained eye for what? I'd assume that everyone judging that photo contest was a professional with a trained eye for photography; but that apparently doesn't imply having a trained eye for ML-generated images, and doesn't necessarily even involve having looked at ML-generated images, much less knowing which are the common "tells" for detecting them.
The person who generated the image seems to have decades of award winning experience in actual photography, so I wonder if that trained eye passes your test?
Calling out your need to put down others with petty, frivolous details to feel better about youself and your insecurities/frustrations is not an issue of offense. You're contributing to shut down an otherwise interesting discussion just because you feel the need to elevate yourself at the expense of others.
> But this is obvious for a trained eye, which a professional must have.
The fact is that you're clearly wrong, as demonstrated by this contest and the fact that the AI generated image was awarded the top prize by its jury.
Your only counterclaim is that you portray yourself as better than others without reason or substance, and in the process tried to make it about you and elevate yourself at the expense of others by spreading baseless accusations of incompetence.
Again, this says more about you than AI generated content and those who evaluate it, and this discussion is really not about you.
Not sure why you’re getting downvotes, your post is completely spot on.
That a professional jury awarded the prize to an AI-generated image is proof that AI has reached a good enough level to fool a trained eye.
To come later, with prior knowledge that it was AI-generated, point out “tells” and generally armchair-quarterback the process, proves absolutely nothing about the jury’s competence.
The jury https://www.worldphoto.org/2023-judges each member more than qualifies as a professional with a trained eye which means your assumption that it is obvious to a professional trained eye is provably false. I deal with AI generated images a lot and have seen a lot of these kinds of finger distortions in AI generated images as well, but with this creative category and the distorted william hope-esque style it's easy to attribute the distortions to other effects if you don't already know or assume it's AI generated.
>Throwing blanket accusations of incompetence at the jury
Clearly you haven't spent much time in the art world. The idea of competence is fundamentally opposed with the reigning ideology in fine art (as opposed to commercial art).
Hard disagree, it's exceptionally obvious - the hands aren't just a little off, they're anatomically incorrect. And this is specifically a competition so one would assume that the judges would have at least minimal aesthetic training.
> Hard disagree, it's exceptionally obvious - the hands aren't just a little off, they're anatomically incorrect.
It really isn't. It's fine you tried to play Catch the Differences after knowing what to look at. That's not what people do with photos running in contests.
Composition push-ups, color-theory pull-ups, subject lighting crunches. It’s not easy but you gotta get artistically ripped before you can be a judge art.
There are steps in solutions like this (I would imagine a first being all images are hashed / signed by a key stored on the camera itself, so the orignal raw image can be audited later - this is something that will help more jounralistically than artists but it's relevant)
But ultimately photography becomes a sport - yes it's possible to run faster / lift more if you take the drugs. But we ban that because of reasons, much like I suspect competitions will ban AI content and possibly so will newspapers.
(which is why the above signing of images might be a huge step forward)
How do you tell that some key is stored solely on a particular camera?
I mean, a camera manufacturer could issue a digitally signed certificate that this key should be only embedded in a particular camera, but it doesn't prove that a copy wasn't made beforehand or extracted from that camera, or that someone wasn't able to misuse the manufacturer's signing keys to sign a few extra keys to be sold to e.g. propaganda manufacturers.
All of that is fair. The secure enclave on an iPhone is only so secure.
But given that any idiot can fake a picture of Obama \ Trump / shaking hands with <insert here> it would be nice to have something that easily and automatically gets repudiated 99% of the time - we want to default trust not default distrust. Life is much much harder the other way round and design society to promote what makes society stronger is a good thing.
In the article, he explains his goal was to accelerate the conversation around AI in photography that the field isn't ready to have yet, and didn't want his submission to win.
It reminded me of the pained actors in the series Berlin, where they are making a movie and and all being overly artistic. Slightly fitting too, though that was set in the ‘20s from memory.
This may not be photography and it’s certainly controversial given the context, but it’s still interesting, meaningful and “good” art produced by an artist… using AI.
Not surprising. Decent AI generated images tend to be better then those created by your typical artist, and AI image generators in the hands of a good artist can create some incredible work that most human artists will struggle to compete with.
Reminds me of those stories of people on horseback trying to race locomotives.
AI art is still 100% art. If any of us ever consider coding artistic and inventive, then AI art is also art.
I code. I am also an oil painter. I suck at composition. If I ask an AI to layout a scene, is my meatspace painting less art? ( i have not done that just got the idea..)
AI is a tool in the toolbox.
Is someone who buys paint in tubes cheating because they did not grind lapis and mix it with rapeseed oil?
Ironically photography was historically seen as “not art” by many traditionalists. That background is possibly why a new form of art using different tools triggers them - photographic competitions (anthropomorphically) are maybe just a tad sensitive on the topic?
And the artists “performance” at the ceremony is top quality art (humour, satire, insight, and a reflection on our society).
His submission might have won the competition because of his name, if he didn’t use a pseudonym. His name would make them think that the submission was human-made, not AI-generated.
A renowned documentary photographer (e.g. worked for Magnum) I've greatly respected recently announced a NFT project consisting of AI-generated photos on the topic of Cuban refugees [0]
> IMAGERY IS NOT REAL
> 90 Miles is a post-photography AI reporting illustration experiment exploring historical events and realities of Cuban life that have, since 1961, motivated Cubans to cross the 90 miles of ocean separating Havana from Florida.
> For over 25 years, I kept a list of subjects to document but was unable to, mainly due to access or that access was impossible. While working in Cuba from 2014-2016, more subjects were added to the list with the assumption they would likely never be created...
> ...The Cubans who attempt the crossing are incredibly resourceful, as necessary to the journey and as reflected in rafts often assembled from inner tubes, pieces of wood and plastic, household supplies, etc.
Of all the AI-art controversies so far, this one really rankled me. This guy has had an immensely successful career in documentary photography, he's gained fame and prestige based on the veneer of truth and reality that journalism confers to his line of work. And now he's going to torch the integrity of the profession for some NFT profit?
Ironically, he's done a bunch of work documenting Cuban life. So the photos he generates will probably bear some resemblance to reality he's observed. But why not amplify the work of young/undiscovered photographers who do have the chance to document the plight of refugees? Just shocked at how little he cares about trivializing refugees and future aspiring documentarians.
> And now he's going to torch the integrity of the profession for some NFT profit?
I think you are missing the point.
There is no "integrity" in photography outside of the photographer one. Images are and always have been fabrication. Choice of subject, framing, setting, careful curation, the photo is a medium of expression which is not that different from writing something.
The message Christopher Brown is conveying is very clear. It’s not that meaningfully different to create the point of view he wants to share using machines or using a camera.
> the photo is a medium of expression which is not that different from writing something.
Yes, of course photography is an expression for the reasons you mention. When I see a documentary photograph, I know it's not the "whole picture" but just a moment lasting 1/200s and maybe 1 out of 500 photos that a photographer picked out. But I do assume that the moment I'm seeing did actually exist from that perspective, and that there are real limits to how the photographer can control the image. That claim to reality is all that documentary photography has to elevate it from artistic and illustrative photography.
Here's an example (from the NFT AI Cuban project) of what gets lost when we treat realistic AI-generations on the same epistemic level of a traditional documentary photo
Knowing that this photo is AI generated, what are we supposed to make from it? Of course the human pictured doesn't exist, but does this helmet actually exist as part of the Cuban refugee life, or is it purely an artifact of the AI's hallucination? Documentary work has been great at showing how fact is stranger than fiction — e.g. how during in war, people can have moments of laughter and goofiness. Treating AI photos and documentary work on equal footing ultimately erodes our understanding of how expansive and weird reality can be.
There's an important point to consider. Our ability to detect "real" dialogue, images and whatnot will develop as we go. What fools us now will not fool us as easily in the near future.
That doesn't mean technology won't outrun our ability to detect it, just that this images will probably seem like an obvious AI image in a few years.
> Just as photography replaced painting in the reproduction of reality, AI will replace photography. Don’t be afraid of the future. It will just be more obvious that our mind always created the world that makes it suffer.
It kind of did - the vast majority of the jobs done by painters were actually replaced; painting in general and painting as fine art continued on but the vast majority of painting which was done for mundane decorative purposes effectively shrunk away, as did most of the "portraiting" industry; in some cases people still order their portraits (e.g. the gallery of presidents) but is far scarcer and employs far less painters, as for this "job" painting was replaced by photography.
Surely the grotesquely malformed hands was a clue.
For some reason image AI have a terrible time with hands. Some platforms produce the wrong number of fingers for instance, sometimes wildly wrong - 12, 15, 20 fingers.
maybe it is because I spent a lot of time with AI generated imagery but I would figure notice that this is an AI image right away - even ignoring the hands (the days of us being able to do this by eye is numbered though). Surprised that the jurors in a prestigious photography contest did not figure it out or even suspected.
I'd assume that there are many people in the art world who have never seriously looked at AI generated imagery - especially the part of community which is more prestigious, more experienced, older and more traditional, i.e. those most likely to be jurors in contests.
To me, this just seems like a publicity stunt by the artist. Looking at the other images in the category [0], at least three of them are heavily edited. I see no reason why editing his image with AI-assisted tools should automatically disqualify the artist, unless the contest terms prohibit that, of which I see no mention in the article.
What's left is an artist who sets up the contest organizers by
• happily accepting the selection of his work on his website a month before the event (without a mention of intending to refuse the prize),
• then literally the night the before the event, publishing an "open letter" about refusing the prize on the website,
• and finally, showing up uninvited on stage to make his speech about refusing the prize.
This is a clever way of increasing his name awareness by riding the AI art controversy wave, for sure.
> I see no reason why editing his image with AI-assisted tools should automatically disqualify the artist, unless the contest terms prohibit that, of which I see no mention in the article.
He didn't edit an image so much as generate it. His work is closer to a digital painting (from a Wacom tablet, say), than a color-corrected, cropped, then filtered photograph. Everyone else started their process with capturing light from a camera.
> I call my images “images”. They are synthetically produced, using “the photographic” as a visual language. They are not “photographs”.
> Participating in open calls, I want to speed up the process of the Award organisers to become aware of this difference and create separate competitions for AI-generated images.
That's kind of a "tu quoque" analysis of what's going on here.
Anyway. People who use AI editing tools are at least starting with a photograph, or a collection of photographs, presumably. It is now impossible to tell, by simple visual inspection, if something is a photograph or was created from image generation tools. That's historic.
You can tell the image is AI generated by looking at the fingers resting on the shoulder. The woman’s right hand has no fingernails and a claw index finger, her left hand has a wobbly and twisted pinky, a bendy ring finger and a squashed middle finger.
(1) He says specifically that he intended it as a stunt
(2) He didn't edit any image with AI, he generated it from nothing with text.
As it was intended to be a photo competition, he felt that AI images don't deserve to win "photography" competitions.
----
Often the only way to secure anything is to try and break it. This is what we do with penetration testing in technology systems.
The problem is that the organizers aren't being proactive, so sometimes the only next step is some grey-hat stepping in to wave a big flag and saying, "hey, you might want to fix this"
Most forms of protest must by definition cause inconvenience in the short term to create the possibility of positive change in the long term.
Would it have been better if someone had earnestly submitted an AI “photo” and accepted the prize quietly instead?
Reducing this to just a “douche move” seems to miss the forest for the trees. The contest will cease to exist or at minimum cease to have any value/meaning if AI submissions become the norm. An alternate framing is that such a form of protest is a necessary inoculation to prevent infinite “douche moves” going forward.
It’s a douche move because it’s ruining something that is based on trust by betraying that trust. That’s not a challenge, it’s not helping anyone, it’s not high and mighty, it’s just that, a douche move.
Look at how much praise athletes got by proving they could win contests using doping. They got none because it’s a douche move, even if it’s technically not illegal it’s still cheating, even if they never pick up the prize it’s still ruining the contest. Everyone can imagine you can win a race on doping, nobody learned anything.
How will they enforce it though. Will they require you to send a video of creating the work? What will happen when people can fake that too? Pixel forensics?
There may be ways to check the first output of an AI algorithm but there are a lot of filters and simple techniques that can disguise the source of an image.
Giving them an environment in which they don't have access to the AI, I guess: Photos must be signed by the camera firmware, digital art must be made with approved tools on a locked-down device or web VM, etc.
Sorry if I am to lazy to look this up but was this competition specifically for "PHOTOGRAPHS" as the capitalized in his statement? If others were heavily modified ... than this seems to point out that this is a flawed competition rule. So they want AI and Photoshop stuff compete with photographs?
I have a publicity stunt feel but I have to agree at the same time. What is even the point, this leads to photographers being pressured using AI and Photoshop to be "creative" if this is allowed, they should not talk about photographs if that case and make different categories.
Issue is, you can not verify if some slight filters and stuff is used on photos. Not know much about the topic but I guess they all use some kind of filters. But that is still very different than just telling an AI to generate picture for you, even if you feed it your own originals.
Both photography and image generation by computers are techniques that you can use in the ancient discipline of painting, aka "putting colors on a surface to produce a visual effect". Since its invention, photography has been a staple tecnhique for many painters. I don't see why "AI" should be any different.
Leonardo used the camera obscura as part of the creation of some of his paintings. Rockwell worked from photographs to produce most of his illustrations. Many serious artists in the future will certainly spend time crafting prompts.
Honestly this speaks more about their so-called competition for this to have taken first prize, because that "photograph" has some pretty big AI generation red flags in it, including the terribly smudged and distorted hands.
Absolutely. This was the case even when the artist won that art competition using midjourney. The overall mood of the picture was nice but the moment you looked at it a little more closely, everything that wasn’t the immediate subjects kind of warped into nonsense. Would love to hear from the judges of these competitions on how they select winning entries
135 comments
[ 2.3 ms ] story [ 183 ms ] threadIt is now like you had contests for nail smiths, and suddenly you could buy the nicest nails at home depot.
Edit: Also you wouldn't hand out prices for a reprint of an oil painting.
With blackjack. And hookers.
Sure, they should be treated as separate categories of art, just as we don't include photographs in painting contests, however, that was solved by the artist promptly refusing the prize and disclosing the origin; but until this automatically dismissive attitude persists, triggering this discussion is a good thing - and I'd rather condemn the actions of the organizers, attempting to silently erase this event and pretending it didn't happen. Their jury did decide that at least in some cases an AI generated image can be better art than a "manually" taken photo, and they need to own that decision.
Should we "condemn" the judges of chess tournaments for not allowing live coaching and chess AIs?
If the chess community would assert that computers are worse than humans as chess and couldn't ever match human play, and when someone won a chess tournament with forbidden computer assistance the judges of that tournament would instead of saying "yep, the win was due to a forbidden advantage, so we disqualified that participant" rather hide the fact and go "no, that didn't happen, obviously human reasoning won all the games", then we should absolutely condemn them for that hypocrisy.
His work is quite strong.
https://www.eldagsen.com/pseudomnesia/
"Just as photography replaced painting in the reproduction of reality, AI will replace photography. Don’t be afraid of the future. It will just be more obvious that our mind always created the world that makes it suffer."
Rather what people were using photography for in the past may be replaced by AI. Before people went to stock photo sites for cheap images to put on their website or whatever, and tomorrow they'll find out it can be cheaper on midjourney or stable diffusion and it will do that.
But AI won't replace photography for what photography does, rather it will replace photography as the cheapest satisfactory image to use on your website that doesn't need a depiction of reality.
That's one use of photography. OP is about an art contest, not a photojournalism contest.
But as far as "reproduction of reality", obviously AI productions unreality won't replace photography. It might replace reproduction of reality, though.
No, OP is about a photo contest, not an art contest.
>World Photography Organization’s Sony World Photography Awards
AI is even less than that - people who want a computer to make their artistic decisions for them have a neural net author an image or other work for them. generally the humans having no more artistic input than someone who commissions a piece from an artist by explaining what they want.
A person painting makes a trillion times more artistic decision making than someone 'painting' with a neural net. They aren't the author of the work, and can't enter it into art competitions on their own behalf (unless the rules are stupid), and can't register it for copyright as there is no human authorship.
It is the complete dissolution of human art making - its humans deciding to have a non human thing - that cannot feel emotions or understand the emotions in an image - make 'art' for them. Most because they are incapable of doing it themselves so having something is better than having nothing, some because they do art primarily for money and the neural net works faster than them and they haven't figured out the bit about no copyright yet, or will misrepresent it to illegally get a copyright on an image they didn't author.
That last point about feelings and emotions however is not clear cut, imo. Typically the process is itertive, so it's a loop and there is a feeling agent in that loop. Also if you yourself make art, you must have noticed how sometimes even the simplest of tools (say a brush or nib) create ~accidental features of the work that one finds entirely pleasing. This happens a lot to me with ink and brush :) The general input is distincly my line, it is my hand w/o a question, but I have a few works which feature engaging aspects that were partly accidental, and part of the delight is precisely that: me, the supposed author, is surprised.
Is someone who continually commissions art until they arrive at a painting they like making artistic decisions in the work? I don't think so.
>you must have noticed how sometimes even the simplest of tools (say a brush or nib) create ~accidental features of the work that one finds entirely pleasing. This happens a lot to me with ink and brush :) The general input is distincly my line, it is my hand w/o a question, but I have a few works which feature engaging aspects that were partly accidental, and part of the delight is precisely that: me, the supposed author, is surprised.
That is unquestionably something you authored, even the mistakes, even the mistakes you liked. But at no point of commissioning artwork are you making artistic decisions.
It's an accurate description of whats going on. Commissioning paintings until you get the one you want isn't artistic expression, or artistic decision making, and the human coming up with prompts will be replaced by a machine learning program trained on prompts, as they bring the least value to the process and are the most easily replaced.
> And from a purely practical point of view, getting worked up over this is tilting at windmills.
I'm not worked up at all. I'm explaining whats going on. Is anyone who thinks differently from you getting worked up? Are you getting worked up for imagining emotions I'm not experiencing? In the future just argue ideas instead of imagining emotions across internet mind reading.
>Change is the constant. And there will be undoubtedly 'artist commissioners' that will outshine others, using the same 'artist'.
It will all approach zero value, so it doesn't matter. It's inauthentic, uncopyrightable stuff based on the misappropriation of the work of actual artists. The expert 'commissioners' will also be replaced by machine learning faster than anyone else.
The subject matter is probably what "won" the contest, it's pretty well composed.
https://www.eldagsen.com/wp-content/uploads/2023/04/preview_...
.. I was impressed with composition(s) as well.
It’s more like the training data is used to score the generated images against to guide it into producing something that looks similar to a real image
This, seems outside the lines. Entering a photography contest with work that isn't based in photography, that you didn't shoot, isn't in the spirit of the contest. A photograph is an image created by light falling on a photosensitive surface - Edit all you want, hell edit till it's not even the same thing anymore, but if you're entering a photography content, at least trigger a shutter. The Sony photography awards are very clearly based around photography as the fundamental starting point. There are lots of pure digital art contests to enter, so enter them.
[1]https://news.ycombinator.com/item?id=35020010
Maybe "photography" contests should be more strict, with only specific processing filters allowed.
Photos used for editorial purposes in newspapers and magazines at least historically had pretty strict rules about manipulation generally--including the removal of distracting elements that had nothing really to do with the core "truth" of the photo. I suppose the thinking was that manipulation can quickly become a slippery slope.
You're basically putting photos in front of people who've seen it all. They're not impressed by technical perfection or what gets 90+ on 500px.
edit: It's more likely your friend was losing out to people who could take more photos and be more experimental due to not being limited by the the cost of film or delay of development. Film can't touch digital's feedback loop.
In any case, perhaps some kind of hardware-signed cryptography scheme on the files from the camera could be used in lieu of this development in the future.
"Editable" artificial images?
I'll admit those are stretches. Anyway, in addition to a digital signing scheme, these competitions are gonna need more editing restrictions if they want to exclude the use of diffusion models.
RAW files generated "backwards" do already exist. For this, police and forensics have been using special cameras with hardware signatures for raw files for around 15 years, but Canon's and Nikon's system was broken by Elcomsoft like 12 years ago. Sony is trying something similar in their A7 IV, but I don't know how serious it is, and I think it's not as oriented to forensics as Canon and Nikon was.
https://www.elcomsoft.com/news/428.html
https://www.prnewswire.com/news-releases/elcomsoft-claims-to...
https://m.dpreview.com/news/5658217744/sony-a7-iv-gets-anti-...
The author traveled to the ceremony at his own expense, found that he was not allocated time to speak, then went onstage unannounced to explain his refusal, at which point all physical evidence of his work was removed from the facility, again without comment.
I agree that Sony could have saved itself trouble by requiring the raw camera footage with verifiable exif data.
Sony problem was in not catching someone who wanted to embarrass them. There's not a good way to detect that in advance.
Sony makes cameras; signing the image in the exif with a public key would establish authenticity.
This event isn't likely to spur that change, but it may well come.
Feels dystopian that they would just, scrub the winning work from the website and physical location without comment, pretending as if it didn't happen.
the fact that he was able to enter at all with a non-photo is proof that Sony is in the wrong, and despite their coverup, is not, in fact, prepared to deal with "black-hat artists" in this space
when an inadequacy makes itself apparent, the decent thing to do is acknowledge it, thank anyone who helped you see it, and try to make it right – Sony did none of these things, and thus are the "indecent" ones here
- Changing the style of the whole photo to (for instance) make it look old and analog.
- "Enhancing" it with HDR, depth of field, focus everywhere or something like that.
- Altering exposure beyond what you could get out of a underexposed/overexposed RAW file.
- Stylizing the human subjects to make them look more attractive, more grimey or whatever
All these things could arguably be done in photoshop, but its just very labor intensive and more hit-or-miss. Hence the contest rules would have to exclude most photoshop work if they want to exclude diffusion.
There is a big distance between filters which tweak an existing image and generating an image from nothing.
And its literally a continuum from "leave the image 99.9% unchanged" to "change 99.9% of the image"
Even in traditional film photography, you can easily get "untrue" pictures. Film is sensitive to UV and near infrared, so filters can create very unreal imagery. And multiple exposure can do all sorts of tricks. The averaging of many faces together, creating a sort of uncanny valley result, was first done in the analog, stacking aligned photos with multiple exposure.
This is generally still considered photography. Art photography, if it's any good.
One could say the key element is the artist combining actual scenes that were exposed to film or a sensor. Perhaps massaged a bit in the dark room or digitally. But the core material is exposed images from actual scenes. It's photography while that element remains dominant. But then again ... isn't that sort of what these generative AIs also do? It's all originally derived from actual photographs.
Trouble is, his work wasn't img2img, there was no initial image as input
So, on AI generators. I am an artist, I have painted and photographed all my life. The average prompter has no chance to win against a professional. I know which artist is presented in the data set, how to use the light and composition. So basically speaking, this is the new Photoshop. The interesting question is what will hold long term more value in the future? The generated content or analogue product.
The pupils aren’t quite right on the front woman’s eyes. Her right eye has a sort of lump and her left iris isn’t quite round.
The fabric over chest looks more like carved stone.
100% I’d not have said ‘AI’ on first look though, and if someone did a reverse-switcheroo and said ‘actually it’s real’ I could be convinced.
I think you're trying very hard to poke holes at a very run-of-the-mill process. It's easy to pick apart stuff post-facto when you have the benefit of context and hindsight, and you are certain of what you're looking at.
If instead you're judging artwork, it would be unthinkable to reason "this work is pretty good but the model's fingers might look a bit off, and that is a good enough reason to reject the prize".
Throwing blanket accusations of incompetence at the jury also says more about yourself than the contest.
The truth of the matter is that AI -generated images are becoming quite impressive and can be a challenge to tell apart. This is not a problem with a jury in a context. We had deep fakes in the recent past, and they are improving rapidly.
It's easy to try to elevate yourself by putting down others, but while you're there making claims like it's obvious that the streaming audio of Biden and Obama and Trump and Hillary hanging out playing Minecraft is bullshit, multiple high-ranking officials already fell for calls from people using similar tech to impersonate world leaders.
Calling out your need to put down others with petty, frivolous details to feel better about youself and your insecurities/frustrations is not an issue of offense. You're contributing to shut down an otherwise interesting discussion just because you feel the need to elevate yourself at the expense of others.
> But this is obvious for a trained eye, which a professional must have.
The fact is that you're clearly wrong, as demonstrated by this contest and the fact that the AI generated image was awarded the top prize by its jury.
Your only counterclaim is that you portray yourself as better than others without reason or substance, and in the process tried to make it about you and elevate yourself at the expense of others by spreading baseless accusations of incompetence.
Again, this says more about you than AI generated content and those who evaluate it, and this discussion is really not about you.
That a professional jury awarded the prize to an AI-generated image is proof that AI has reached a good enough level to fool a trained eye.
To come later, with prior knowledge that it was AI-generated, point out “tells” and generally armchair-quarterback the process, proves absolutely nothing about the jury’s competence.
Clearly you haven't spent much time in the art world. The idea of competence is fundamentally opposed with the reigning ideology in fine art (as opposed to commercial art).
It really isn't. It's fine you tried to play Catch the Differences after knowing what to look at. That's not what people do with photos running in contests.
Composition push-ups, color-theory pull-ups, subject lighting crunches. It’s not easy but you gotta get artistically ripped before you can be a judge art.
But ultimately photography becomes a sport - yes it's possible to run faster / lift more if you take the drugs. But we ban that because of reasons, much like I suspect competitions will ban AI content and possibly so will newspapers.
(which is why the above signing of images might be a huge step forward)
Better just to stick with the social, and tell people that violating the rules will result in disqualification and a ban.
I mean, a camera manufacturer could issue a digitally signed certificate that this key should be only embedded in a particular camera, but it doesn't prove that a copy wasn't made beforehand or extracted from that camera, or that someone wasn't able to misuse the manufacturer's signing keys to sign a few extra keys to be sold to e.g. propaganda manufacturers.
But given that any idiot can fake a picture of Obama \ Trump / shaking hands with <insert here> it would be nice to have something that easily and automatically gets repudiated 99% of the time - we want to default trust not default distrust. Life is much much harder the other way round and design society to promote what makes society stronger is a good thing.
Did he submit his work in the contest and then refuse the prize ala Sokal [1] because it was generated using AI?
[1] https://en.wikipedia.org/wiki/Sokal_affair
What he didn't want was the prize for a photo contest.
That sentence infuriates me way more than it should. Are you sure it's Latin? Are you sure it's not Swahili? Also, don't define a term with itself.
https://medical-dictionary.thefreedictionary.com/pseudomnesi...
https://dictionary.apa.org/pseudomemory
https://www.dictionary.net/pseudomnesia
Reminds me of those stories of people on horseback trying to race locomotives.
I code. I am also an oil painter. I suck at composition. If I ask an AI to layout a scene, is my meatspace painting less art? ( i have not done that just got the idea..)
AI is a tool in the toolbox.
Is someone who buys paint in tubes cheating because they did not grind lapis and mix it with rapeseed oil?
And the artists “performance” at the ceremony is top quality art (humour, satire, insight, and a reflection on our society).
> IMAGERY IS NOT REAL
> 90 Miles is a post-photography AI reporting illustration experiment exploring historical events and realities of Cuban life that have, since 1961, motivated Cubans to cross the 90 miles of ocean separating Havana from Florida.
> For over 25 years, I kept a list of subjects to document but was unable to, mainly due to access or that access was impossible. While working in Cuba from 2014-2016, more subjects were added to the list with the assumption they would likely never be created...
> ...The Cubans who attempt the crossing are incredibly resourceful, as necessary to the journey and as reflected in rafts often assembled from inner tubes, pieces of wood and plastic, household supplies, etc.
Of all the AI-art controversies so far, this one really rankled me. This guy has had an immensely successful career in documentary photography, he's gained fame and prestige based on the veneer of truth and reality that journalism confers to his line of work. And now he's going to torch the integrity of the profession for some NFT profit?
Ironically, he's done a bunch of work documenting Cuban life. So the photos he generates will probably bear some resemblance to reality he's observed. But why not amplify the work of young/undiscovered photographers who do have the chance to document the plight of refugees? Just shocked at how little he cares about trivializing refugees and future aspiring documentarians.
[0] https://www.instagram.com/p/CqniSVjrj0h/?hl=en
I think you are missing the point.
There is no "integrity" in photography outside of the photographer one. Images are and always have been fabrication. Choice of subject, framing, setting, careful curation, the photo is a medium of expression which is not that different from writing something.
The message Christopher Brown is conveying is very clear. It’s not that meaningfully different to create the point of view he wants to share using machines or using a camera.
Yes, of course photography is an expression for the reasons you mention. When I see a documentary photograph, I know it's not the "whole picture" but just a moment lasting 1/200s and maybe 1 out of 500 photos that a photographer picked out. But I do assume that the moment I'm seeing did actually exist from that perspective, and that there are real limits to how the photographer can control the image. That claim to reality is all that documentary photography has to elevate it from artistic and illustrative photography.
Here's an example (from the NFT AI Cuban project) of what gets lost when we treat realistic AI-generations on the same epistemic level of a traditional documentary photo
https://twitter.com/SaabGT750/status/1643651637033017345
Knowing that this photo is AI generated, what are we supposed to make from it? Of course the human pictured doesn't exist, but does this helmet actually exist as part of the Cuban refugee life, or is it purely an artifact of the AI's hallucination? Documentary work has been great at showing how fact is stranger than fiction — e.g. how during in war, people can have moments of laughter and goofiness. Treating AI photos and documentary work on equal footing ultimately erodes our understanding of how expansive and weird reality can be.
That doesn't mean technology won't outrun our ability to detect it, just that this images will probably seem like an obvious AI image in a few years.
For now, we need to manually count knuckles.
That is a pretty amazing quote.
For some reason image AI have a terrible time with hands. Some platforms produce the wrong number of fingers for instance, sometimes wildly wrong - 12, 15, 20 fingers.
What's left is an artist who sets up the contest organizers by
• happily accepting the selection of his work on his website a month before the event (without a mention of intending to refuse the prize),
• then literally the night the before the event, publishing an "open letter" about refusing the prize on the website,
• and finally, showing up uninvited on stage to make his speech about refusing the prize.
This is a clever way of increasing his name awareness by riding the AI art controversy wave, for sure.
[0] https://archive.is/pH3Jb
He didn't edit an image so much as generate it. His work is closer to a digital painting (from a Wacom tablet, say), than a color-corrected, cropped, then filtered photograph. Everyone else started their process with capturing light from a camera.
> I call my images “images”. They are synthetically produced, using “the photographic” as a visual language. They are not “photographs”.
> Participating in open calls, I want to speed up the process of the Award organisers to become aware of this difference and create separate competitions for AI-generated images.
https://www.eldagsen.com/sony-world-photography-awards-2023/
Anyway. People who use AI editing tools are at least starting with a photograph, or a collection of photographs, presumably. It is now impossible to tell, by simple visual inspection, if something is a photograph or was created from image generation tools. That's historic.
(1) He says specifically that he intended it as a stunt
(2) He didn't edit any image with AI, he generated it from nothing with text.
As it was intended to be a photo competition, he felt that AI images don't deserve to win "photography" competitions.
----
Often the only way to secure anything is to try and break it. This is what we do with penetration testing in technology systems.
The problem is that the organizers aren't being proactive, so sometimes the only next step is some grey-hat stepping in to wave a big flag and saying, "hey, you might want to fix this"
The point of a stunt like this is to prove that a non-photo can win a photography contest, and to highlight how absurd that is.
Would it have been better if someone had earnestly submitted an AI “photo” and accepted the prize quietly instead?
Reducing this to just a “douche move” seems to miss the forest for the trees. The contest will cease to exist or at minimum cease to have any value/meaning if AI submissions become the norm. An alternate framing is that such a form of protest is a necessary inoculation to prevent infinite “douche moves” going forward.
Look at how much praise athletes got by proving they could win contests using doping. They got none because it’s a douche move, even if it’s technically not illegal it’s still cheating, even if they never pick up the prize it’s still ruining the contest. Everyone can imagine you can win a race on doping, nobody learned anything.
Where and how the line would be drawn and you'd enforce it would be interesting discussions though.
There may be ways to check the first output of an AI algorithm but there are a lot of filters and simple techniques that can disguise the source of an image.
I have a publicity stunt feel but I have to agree at the same time. What is even the point, this leads to photographers being pressured using AI and Photoshop to be "creative" if this is allowed, they should not talk about photographs if that case and make different categories.
Issue is, you can not verify if some slight filters and stuff is used on photos. Not know much about the topic but I guess they all use some kind of filters. But that is still very different than just telling an AI to generate picture for you, even if you feed it your own originals.
Leonardo used the camera obscura as part of the creation of some of his paintings. Rockwell worked from photographs to produce most of his illustrations. Many serious artists in the future will certainly spend time crafting prompts.