Welcome to DU!
The truly grassroots left-of-center political community where regular people, not algorithms, drive the discussions and set the standards.
Join the community:
Create a free account
Support DU (and get rid of ads!):
Become a Star Member
Latest Breaking News
Editorials & Other Articles
General Discussion
The DU Lounge
All Forums
Issue Forums
Culture Forums
Alliance Forums
Region Forums
Support Forums
Help & Search
See The Bizarre Image That Just Won an Inaugural Award For AI Art
Excerpts from the Forbes piece and then an Australian news site below, followed by my comments. I consider this sort of thing not just creepy but harmful, as I'll explain in reference to some of what's quoted in both articles.
https://www.forbes.com/sites/lesliekatz/2023/10/08/see-the-surreal-image-that-just-won-an-ai-art-contest/amp/
In the black-and-white image, two women tightly cradle each other–and an octopus. Humans don't hug cephalopods every day, so how did photographer Annika Nordenskiöld manage to capture the highly unusual sight? You guessed it. With the help of AI.
The lifelike picture, titled “Twin Sisters in Love,” on Saturday won the inaugural Prompted Peculiar International AI Prize at the Ballarat International Foto Biennale, an Australian photography festival running through October 22. The competition is believed to be one of the first, if not the first, AI-art award.
For the winning image, Nordenskiöld, who lives and works in Sweden, partnered with Midjourney, an AI tool that quickly turns text phrases, or “prompts,” into hyper realistic images by scanning a massive database trained on visual art by humans. Artificial intelligence tools like Midjourney, Dall-E and Stable Diffusion continue to capture imaginations, as they let anyone generate images from text in mesmerizing and sometimes creepy and wildly absurd ways.
“None of the places, people or creatures in my prompts exist in the physical realm,” Nordenskiöld said of her winning creation in a statement. “They were conjured from the sum of human experience in our deep collective well, as seen from my dreamboat with its flickering light.”
-snip-
The lifelike picture, titled “Twin Sisters in Love,” on Saturday won the inaugural Prompted Peculiar International AI Prize at the Ballarat International Foto Biennale, an Australian photography festival running through October 22. The competition is believed to be one of the first, if not the first, AI-art award.
For the winning image, Nordenskiöld, who lives and works in Sweden, partnered with Midjourney, an AI tool that quickly turns text phrases, or “prompts,” into hyper realistic images by scanning a massive database trained on visual art by humans. Artificial intelligence tools like Midjourney, Dall-E and Stable Diffusion continue to capture imaginations, as they let anyone generate images from text in mesmerizing and sometimes creepy and wildly absurd ways.
“None of the places, people or creatures in my prompts exist in the physical realm,” Nordenskiöld said of her winning creation in a statement. “They were conjured from the sum of human experience in our deep collective well, as seen from my dreamboat with its flickering light.”
-snip-
From Australia's Sydney Morning Herald:
https://www.smh.com.au/culture/art-and-design/world-s-first-ai-art-award-ignites-debate-about-what-is-photography-20231004-p5e9td.html
The organisers of the Ballarat festival hope the inaugural AI prize, Prompted Peculiar, a world first for an international photography festival, will open debate about the difference between photography and what they call “promptography.”
The term was coined this year when German photographer and judge at the Ballarat festival Boris Eldagsen turned down first prize in the creative open category of the Sony World Photography Awards because his image “The Electrician” was generated by AI, and in his words, was “not photography”.
“At the time, I was considered a fire starter in the photography world in opening up this conversation on ‘promptography’, but now I am seen as the Che Guevara of analogue photographers and how we approach the elephant in the room – AI-generated images,” he said.
“Now the default position for when you look at pictures online will be to assume they are artificially generated – rarely are they fact-checked ... which presents a problem for picture editors, photographers, developers, AI experts and social media consumers the world over.”
The term was coined this year when German photographer and judge at the Ballarat festival Boris Eldagsen turned down first prize in the creative open category of the Sony World Photography Awards because his image “The Electrician” was generated by AI, and in his words, was “not photography”.
“At the time, I was considered a fire starter in the photography world in opening up this conversation on ‘promptography’, but now I am seen as the Che Guevara of analogue photographers and how we approach the elephant in the room – AI-generated images,” he said.
“Now the default position for when you look at pictures online will be to assume they are artificially generated – rarely are they fact-checked ... which presents a problem for picture editors, photographers, developers, AI experts and social media consumers the world over.”
First of all - Eldagsen is absolutely right about photo-like AI images being "a problem for picture editors, photographers, developers, AI experts and social media consumers the world over."
But he minimizes it. They're a disaster. Not just for artists, but for society in general, including governments, especially democracies, because they make it possible for AI users to come up with potentially convincing deepfakes in under one minute (type in a brief prompt and the AI image generator will typically offer several different image options - I think most offer 4 options - in under 10 seconds). These images' existence can make people skeptical of even real photos they see. These AI deepfakes' existence is now being used by defense attorneys to tell juries not to believe their eyes. They undercut democracies and justice systems.
Second of all, there is no such thing as promptography. No such thing as a promptographer. It's a ridiculously pompous way to refer to giving a text prompt to an AI image generator, something that anyone, including any child, can do. Compared to the control photographers have and the decisions they have to make to get great photographs, "promptographers" are blindfolded in straitjackets strapped into canoes sent down rapids and over waterfalls in an AI river. God only knows where or how they'll end up, but they certainly don't. And neither does the AI, which is mindlessly following algorithms as it meanders through vast data sets of stolen images. Anything a "promptographer" puts in an image prompt does nothing more than the equivalent of trying to specify the location on that river in the hope the ride might land them somewhere safe and recognizable. But it doesn't always. I posted recently about an image generator tested with a prompt requesting an image of a ballerina in an arabesque position. Of the 4 options generated, only 1 showed that position - and that one had the ballerina's leg painfully if not impossibly contorted. Promptographers discard most of what AI offers and have real artistic control over none of it.
Finally, I can't ignore what the winning fake photographer said here:
“They were conjured from the sum of human experience in our deep collective well, as seen from my dreamboat with its flickering light.”
That's not only sounds very sadly conceited, but it's BS.
The AI mindlessly throws together images following algorithms through countless stolen images. Sometimes watermarks that the creators of those original photos had hoped would protect their work will pop up and have to be erased with an editing tool. And there is no "dreamboat with its flickering light." It's software anyone can use, if they don't mind using unethical software that could very well offer the same or a nearly identical image to anyone anywhere at any time, because the AI has no awareness of what it's spitting out.
The prompt for this ridiculous AI image was probably something like: "Photorealistic black-and-white image of twin sisters in polkadot blouses hugging a pet octopus they love."
And by specifying black and white, the odds were increased of getting retro clothing and hairstyles.
Any number of images might have been discarded before this one was churned out. And maybe the prompt was tweaked a bit.
But that was all the "artistry" required, thanks to the theft of all those photos for Midjourney's data set.
The woman who used AI to churn out this fake photo may well be a fine artist otherwise, and she deserves applause for her real art.
But this isn't art. Any more than a 5 year old typing "puppy" as a Midjourney prompt would be an artist.
21 replies
= new reply since forum marked as read
= new reply since forum marked as read
And the flock follows. 
Oh, god! I'm getting that engraved on a brass plaque.
The future just keeps on sucking harder.
Nailed it.
(Richard Avedon)
(composite photo by Flora Borsi)
Neither of these were created with generative artificial intelligence. Both would likely strike many viewers as bizarre - and unnatural. Both artists have received critical acclaim for their work.
It seems to me that this particular contest would be a good place to start a good, substantive, discussion of the place of AI in image-making.
The contest was expressly for images created using AI. So the images entered in the competition are not "deepfakes." The competition was created after an individual (Boris Eldagsen) won a photography competition with an image he created using artificial intelligence - one which he told those running the competition was created with artificial intelligence before they identified him as the winner. He declined the honor because he did not believe his image should have won a photography competition. Partly in response, this competition was expressly for images created using AI.
And, to further distinguish these images from photographs (and his work from that of photographers), he and others coined the term promptographer to describe what they do. Unfortunately, rather than celebrate the fact that they are doing exactly what you have been urging people on DU to do - to be up front about how the images were created - you are dismissing it as "no such thing." At one point photography didn't exist, nor did the term photographer. At some point after it did, the term photographer was created to designate those who practice it. There is a simiar process going on here. Promptographer may not stick - but it's a place to start.
I don't know how this promptographer created her image - I have been unable to find a detailed description. But I do know how the earlier promptographer created his image. It was far more than drafting a few words and picking which generated image he liked. He selected an image, identified parts of the image he didn't like, discarded those and generated a new sub-image to fill those parts. He also expanded his canvas beyond the boundaries of the initial image and filled those as well. It was an iterative process that required him to use an artist's eye to identify when he had the components he believed made a pleasing image. The process, for three images, took 80 hours.
Realism is just one form of art. Just because an image is bizarre - or not photo-realistic does not make it something other than art.
You may not believe images created using AI are art, but academic BFA programs would differ with you. I am currently pursuing a BFA in photography. We are using generative AI in our commercial photography class, and I expect we will using it in other photography classes as well. Because I started out taking a class here and there, I am now backtracking and taking the broad cross-section of classes required for the fine arts portion of the BFA. The use of AI was discussed in the broad overview class every BFA student is required to take as a new tool that we will learn to use in several of our classes. It's not just the local univeristy where I am enrolled. Look up the recognized names in art education (SCAD, Parsons, etc.) and you will find they are also embracing the use of AI as a tool.
Certainly there need to be ethical principles created. In general art is not represented as something it is not. This is often accomplished in the artist's statement which identifies the media. This helps the viewer distinguish between - for example - photorealistic drawings and photographs. It will also likely come to be used similarly for images created using AI.
This image which won was entered into a contest by a person who describes themselves as a promptographer - the very kind of practices you have been urging people on DU to adopt. Why use someone following those practices to rant about the danger of "deepfakes" - which inherently implies they are pretending their image is something it is not? This particular image was likely not generated in a single step from a prompt of a few words. It is far more likely it took 20-30 hours (like the one created by Egaldson). Why use it to suggest the process is a trivial one?
Finally, the software involved in generative AI is not inherently unethical. Software is a tool, nothing more. Training the software on images protected by copyright for which the person providing the training tools does not have a license is unethical (and illegal). But the software itself does not inherently require theft of art. So when you use a phrase like, "It's software anyone can use, if they don't mind using unethical software" it is a "when did you stop beating your mother?" kind of comment. It treats as unethical both those using AI trained on licensed databases and those AI trained on stolen images and builds on that premise, in the same way, "When did you stop beating your mother?" takes, as a premise, that you are actually beating your mother.
Personally - I don't think tools which use unlicensed databases are ethical. I hope the legal system catches up with these tools so that we can have a discussion about AI that is not tainted by how the tools are trained.
Beyond that - I have been making composite photos (like the Borsi one above) for decades in Paintshop Pro to do facial reconstruction by combining intact photos of the person (or relatives they resemble) with a damaged photo. I have been hand-painting in missing details - as well as using clone stamps (a manual version of generative fill) for a similar length of time. As long as I don't pretend it was an unretouched image, I don't see any inherent difference between using generative fill and using the hand techniques I've been using for years. Any more than I see any inherent difference between processing a photo in the darkroom to tailor its appearance - versus tailoring it in an electronic darkroom. It is my artistic judgment that matters, not whether I scratched it out with a stick in the sand or used iterative generative AI to create the image.
There will always be a amateurs who snap photos with their phones and occasionally get one that rivals those of professional photographers - and professional photographers who use cell phones as their tools. I suspect the same will happen with AI. Most of it will be cheap garbage. Occasionally a low-effort AI image will hit it out of the park. But most of the time it will take time, energy, and an artist's eye to use AI to create quality art. Bottom line - it is the artist (and the viewer) which determine if something is art - regardless of the tool(s) used to create it.
Does the availability of new, easier to use, tools mean some artists may no longer be able to make a living at it? Probably. In the same way any advance in technology shifts at least some jobs toward the new technology. We have very few blacksmiths, leech collectors, or human computers anymore.
The fact that alters the jobs available doesn't inherently mean the technology should be rejected.
