Artificial Intelligence Images

[billandsue]

I'm interested in hearing from those of you who have some experience with images created with Artificial Intelligence programs such as Midjourney, DALL-E,  Adobe Firefly, and others.  Specifically, I would like to know what effect these generative AI programs will have on the future of photography.  I have been experimenting with a couple of AI programs, and have attached below an image generated by Midjourney AI following my prompt to create a highly detailed photographic image of two Macaws in a tropical rainforest.  To be honest, I was stunned by the quality of the image, and would love to hear your thoughts.  I would also love to see any AI images you have created.

[pierces]

I've not tried any of the generative AI imaging. I am technically interested in the science behind it but don't see it as something I would use much. Maybe for some illustration work I occasionally do for my daughter when she needs a poster or something for her class. 

 

I'm really not for or against its use. To me, it's just another creative tool used to make art. I would not in any way be in favor of it being used to fraudulently claim the results as an actual photograph, but when identified as an AI-assisted creation, it is art. 

 

I have used AI-assisted programs like Luminar Neo and Topaz Photo AI to make corrections or enlargements of existing photos and have been impressed with the results. They are still in a sort of infancy, but each release produces better and better results.

 

Good discussion to have.

 

Dave

[ski ww]

That is hard to believe it's not a photograph. So you don't know what is real any more.

[Host Bonjour]

On 6/25/2023 at 8:57 PM, billandsue said:

image of two Macaws in a tropical rainforest.

 

There are two birds, their colors are brilliant and it appears to be a rainforest, so the program hit all your marks. Seems to have done a fair job with the feathers, though a few look to be ruffled so maybe it's windy?

 

What I first noticed after the colors, however was the syrupy rain falling from the wood branch, and from there I felt like since I knew it was AI generated, felt challenged to see how well the program did with other photographic details and realism. 

 

I believe it's been reported or understood that AI may be challenged with certain features and eyes may be one of them, so that may be why it's difficult to discern whether the birds' eyes are open or closed...detail is not great. 

 

It also seems that the focus varies, as does the light, though it's possible the forest canopy may vary, from right to left. There also seems to be different weather on the right side of the image than on the left side of the image. With that said, in tropical climates, I know it can rain on one side of the street and not the other, with the sun out, never mind AI.

 

But the depth of field would be consistent from right to left and I'm not sure it looks that way in this image because the fern is more visible behind the bird on the left than whatever might be behind the bird on the right–which could be partly due to the lighting, but a good camera properly set up would have enough light to produce detail on the right side of the image. 

 

Then there's the birds' feet. Again here, the bird on the left fares a bit better and so when I zoomed in on the image to get a better look, it was interesting to see traces of bird skin in places above and below the branch...it seems the program decided to do some camo with the bird's foot and wood. One of the ways in which more detail/refinement is needed.

 

Maybe in the rainforest there is a way rainwater can get syrupy? I'm not sure...but that's what got me into looking. Did the program create what was asked of it? Yes. Would I use it in an ad or magazine or frame it? No. 

 

I'm not sure why yet AI imagery is needed, especially since there are stock photo websites, phenomenal, easy to use photo editing app and programs, smartphones with incredible cameras and better than ever traditional cameras. If you want or need a photo, either it's out there for free (don't take it if it's not clearly free) or to buy or you can make it yourself. 

 

People use actual images to create new things, combining other elements like graphic design or paint, text, and any other of a number of things to create a visual representation of something they'd like to see that for them, is true, unique and one of a kind. 

 

The alarming thing with AI and imagery is that in order for it to generate output, it has to search through images that have already been created by people and use them without their consent to produce an AI image: a new work, which may or may not resemble any of the images harvested in the sourcing for the new output, but which nevertheless could not exist without the creation of prior work. 

 

So, is everyone's work worthy of payment? Maybe not, because not everyone works as a photographer, artist, photojournalist, etc. But everyone does put effort into creating their images and has a personal connection to their photographs, and just because their photos live on the internet, may not want them being usurped for the purposes of random image generation by artificial intelligence. 

 

There had been talk of the ability to include a sort of scrambler or protection filter to the images owned by someone online to help counter the harvesting ability of AI programs, but there were doubts as to the effectiveness and to the impact of the ability to store, host and maintain images on websites/blogs etc when attaching such protections. (Anything added to websites carries a cost and technical element.)

 

AI has enhanced so much of what we do everyday in positive and questionable ways (certain algorithms, ads) so this is an area to keep an eye on. What are the benefits, if any? If we look to the past, some inventions were found to be game changers, some were clever or looked good, but proved to be not worth it in the end. 

A recent Supreme Court case might be a factor in how the use of at least certain images will play out, Goldsmith v. Warhol Estate, which looked at image use rights in whether or not an image was substantially transformed from its original appearance when made into a new work of art that the work was transformed enough so that a usage fee would not need to be paid because the original work was not obvious and recognizable to the viewer of the newly transformed work.

 

Warhol Estate lost the case, because an image of Prince by Lynne Goldsmith was painted by Andy Warhol, then Vanity Fair put the Warhol on its cover and only paid the Warhol Estate, but not Goldsmith. Goldsmith sued...and has won because the Court found that the painting had not substantially transformed her original image. (If Vanity Fair had just paid her, maybe it never would have gone to court; Warhol paid to use the image when first making the painting.)

 

And yes, it's controversial because how is the Court equipped to decide about transformative art etc etc... but the case centered more on Fair Use, an area of copyright law wherein original work that is protected from being reproduced allows limited exceptions for use for free, but the court case obviously was a case involving money. And that's where a lot of the concern arises with AI.

 

Will it put photographers out of work? Will their work be used without permission? How can protections be put in place or guardrails installed on the use of the programs. Should any limits exist?

 

I love what tech has done for creativity, but hopefully it won't obliterate creativity, privacy, or for some folks, rights. We'll see.