https://www.theguardian.com/technology/2023/sep/26/techscape-ai-images-elections-integrity-tiananmen-square

"A strange thing happened last week when you searched for �tank man� on Google.

Tap on image results and instead of the usual photos of Tiananmen Square in Beijing, and the iconic image of a brave protester staring down a convoy of tanks that was captured in 1989, the first result was the same historic moment � but from a different point of view.

For a time last week, the first result on Google Images for �tank man� was instead an AI-generated image of the same protester, taking a selfie in front of the tank. The image was created by Midjourney, and was at least six months old. First reported by 404 Media, a new tech journalism startup set up by former Vice News staff, the emergence of the tank man selfie � which Google subsequently removed from search results for �tank man� � highlighted one of the main fears that Eddie Perez, Twitter�s former head of election integrity, highlighted to me in a recent podcast interview: it�s now possible, with the use of AI imagery, to create alternative history. "

WFT???

https://stock.adobe.com/search?creator_id=210680721&order=creation&k=inisial

They're not the letters the titles say; over 100 meaningless white lines on black. I'd fire any designer that brought me those as logos.

How does this get approved?

If this were not an AI generated image, wouldn't that have been rejected for "aesthetic or commercial appeal of image"?

Yes

https://contributors.gettyimages.com/forum/

You have to be logged in to read it

https://investors.gettyimages.com/news-releases/news-release-details/getty-images-launches-commercially-safe-generative-ai-offering
https://www.gettyimages.com/ai/generation/about

"Customers creating and downloading visuals through the tool will receive Getty Images� standard royalty-free license, which includes representations and warranties, uncapped indemnification, and the right to perpetual, worldwide, nonexclusive use in all media. Content generated through the tool will not be added into existing Getty Images and iStock content libraries for others to license. Further, contributors will be compensated for any inclusion of their content in the training set."

https://www.theverge.com/2023/9/25/23884679/getty-ai-generative-image-platform-launch

https://arstechnica.com/ai/2023/09/getty-images-subscribers-to-get-access-to-ai-image-generator/

https://gizmodo.com/getty-new-art-generator-trained-on-contributors-1850866540

"On an annual recurring basis, we will share in the revenues generated from the Getty Images AI Generator with contributors whose content was used to train the AI Generator, allocating both a pro rata share in respect of every file and allocating a share based on traditional licensing revenue. The first payment is expected to be in late 2024 for the year October 2023-September 2024.

We expect this to represent a new revenue stream for contributors that is additive to the licensing benefits you already enjoy with Getty Images."

From the FAQ:

"What are Getty Images� rights to use my/contributor content for AI training?
Our contributor agreements enable Getty Images to license your content in a broad range of uses, existing or emerging, including training data for AI and machine learning uses."

https://apnews.com/article/getty-images-artificial-intelligence-ai-image-generator-stable-diffusion-a98eeaaeb2bf13c5e8874ceb6a8ce196

The Verge article talks about the quality of the results - they got to try it out - and their screen shot looks miles better than Shutterstock's DALL-E 2 equivalent, at least based on what shows up in the generative AI collection on SS.

"I got a hands-on look at Generative AI by Getty Images and got to play around with the tool for a bit. I mainly wanted to see how it generates photos, rather than illustrations, to test out how close to an actual Getty-watermarked picture it can get. And the photos look better than expected. Stock photos already have an artificial, soulless quality to them, and I was not surprised that some of the first few images the tool generated also felt... devoid of feeling. ...

Getty�s tool did well at rendering realistic-feeling human figures. I prompted it to create a photo of a ballerina in an arabesque position (standing on one leg with the other lifted behind) on a stage with a slightly blurred background. The photos I got felt more human than when I tried the same prompt with Stable Diffusion, and the Getty image fooled my friends when I texted it to them. It's clear Getty�s model trained not just on illustrated art but on actual photos. "

https://techcrunch.com/2023/09/25/getty-images-launches-an-ai-powered-image-generator/
https://www.axios.com/2023/09/25/getty-images-ai-creation-tool
https://fortune.com/2023/09/25/getty-images-launches-ai-image-generator-1-8-trillion-lawsuit/

"The difference, said Getty Images CEO Craig Peters, is this new service is �commercially viable� for business clients and �wasn�t trained on the open internet with stolen imagery.�
He contrasted that with some of the first movers in AI-generated imagery, such as OpenAI�s DALL-E, Midjourney and Stability AI, maker of Stable Diffusion.
�We have issues with those services, how they were built, what they were built upon, how they respect creator rights or not, and how they actually feed into deepfakes and other things like that,� Peters said in an interview."

https://petapixel.com/2023/09/25/getty-images-makes-u-turn-as-it-launches-its-own-ai-image-generator/
(emphasis mine) - I like that way of putting the difference between generative AI and photographs
"However, Getty Images will not allow the material made on its new generative AI tool into its content libraries which will be reserved for �real people doing real things in real places.� "

https://www.business-standard.com/world-news/getty-images-working-with-nvidia-to-debut-its-own-ai-image-generator-123092501302_1.html
https://www.engadget.com/getty-is-going-to-offer-ai-generated-images-after-all-140138829.html
https://www.zdnet.com/article/can-microsoft-recover-from-the-collapse-of-its-surface-business/

Edited Sep 26 to add a few more links to press coverage of this announcement

https://decrypt.co/198660/getty-images-launches-safe-generative-ai-image-tool

From the context, "user-generated" is referring to Getty's customers for their AI tool, not contributors to their "pre-shot" collection

"Getty Images says user-generated images and prompts will train its AI models. Still, as Peters explained, user-generated images will not be uploaded to the Getty Images website or licensed by the company.

�So what you generate and the corresponding outputs are yours to decide whether you want to use or not, but we are not bringing those images back into what we call our pre-shot catalog,� Peters said. �And we don't accept AI-generated images into our pre-shot catalog because we don't know the provenance of what it was created with.� "

https://www.technologyreview.com/2023/09/25/1080231/getty-images-promises-its-new-ai-doesnt-contain-copyrighted-art/

"Tech companies claim that AI models are complex and can�t be built without copyrighted content and point out that artists can opt out of AI models, but Peters calls those arguments �crap.�

�I think there are some really sincere people that are actually being thoughtful about this,� he says. �But I also think there�s some hooligans that just want to go for that gold rush.� "

https://www.forbes.com/sites/johanmoreno/2023/09/25/getty-images-debuts-generative-ai-solution-for-copyright-safe-image-generation/

https://digiday.com/media/getty-images-gets-into-the-generative-ai-race-with-its-own-image-platform/

"�There�s technology for technology�s sake, there�s break things and move fast and ignore other people�s rights � and this doesn�t do [any of] that,� Getty Images CEO Craig Peters told Digiday. �It presents a real meaningful, high-quality solution to customers, which is what they�ve been asking for.�

Similar to competitors like Shutterstock and Adobe, Getty Images offers full indemnification for commercial use of AI-generated images. However, unlike some others, Getty�s AI model is trained on only its own licensed content � a selling point for anyone worried about the range of copyright concerns that plague some other AI platforms.

Customers want to embrace generative AI without having to �absorb a massive amount of IP risk in doing that,� Peters said. He added that the plan isn�t to replace human contributors, but rather to �index on creativity� with �another tool in the creator�s toolbox.�"

https://www.artnews.com/art-news/news/getty-releases-ai-image-maker-trained-on-own-data-1234680408/
https://www.digitalmusicnews.com/2023/09/25/getty-images-generative-ai-platform-music-speculation/

"The concept of unleashing machine learning on a controlled copyright collection is interesting � with a very similar concept rumbling in the music industry. That music-focused model is expected to debut within several weeks, with DMN prepping the story now (stay tuned)."

I can't offer anything specific, but so far, the people who have been blocked had accepted work that fell into that category. So Adobe Stock accepted it by mistake and then blocked the contributor afterwards.

Your situation sounds better in that they rejected it.

What I would be worried about is submitting anything new that fell foul of whatever rule you broke. I can imagine them having rules about repeated violations - although I don't know of anything specific.

Are you now clear on what the violation was and how to avoid it in the future?

...When cameras first came out it was denied any creative value and photographers were not considered artists.
...

This isn't about what is and isn't art. It's about copyright - and in particular US copyright law. Two different issues.

This summary is pretty helpful in outlining what qualifies a photograph for copyright protection.

This article is also helpful in covering the creation of derivative works from a copyrighted photograph (a right that the original owner possesses).

" In photography, whenever someone produces a photograph that closely resembles or is "substantially similar" to another copyrighted work, they may be infringing on the original owner's rights.

To determine if copyright laws have been violated, a court of law compares and evaluates the derivative work against the original. Essentially, a photographer who meticulously recreates an original work's composition, lighting, and other creative aspects is more likely to be found guilty of copyright infringement than a photographer who merely captures subjects that already exist in other photos (e.g., monuments, nature). This allows multiple photographers to photograph the same subjects, like the Golden Gate Bridge, without infringing on each other's artistic rights."

This article talks about the fair use defense - often trotted out by the companies who scraped data without asking for permission or paying.

"Another important fair use factor is whether your use deprives the copyright owner of income or undermines a new or potential market for the copyrighted work. Depriving a copyright owner of income is very likely to trigger a lawsuit. This is true even if you are not competing directly with the original work."

If you read the section on joint authorship in the first article, IMO that comes closest to what's happening with humans using genAI tools. The humans probably wouldn't like that approach as they'd be sharing their copyright with Midjourney or whoever, not to mention it's not clear that Midjourney came by their contribution legitimately:

"Jointly authored works are works that are prepared by two or more creators with the intent that their contributions be inseparable from one another. In a joint work, each of the authors hold an undivided, equal share to the copyright in the final work. Thus, the default rule is that each author has an equal claim to all the exclusive rights in the joint work�the right to reproduce, distribute, create derivative works, and publicly perform or display�unless they separately agree among themselves that the shares or division of rights or profits should somehow be different."

Why not contact iStock and show them how easy it is to replace - perhaps they will come up with a better watermark that can't simply be replaced with generative fill, and/or cropping it?

iStock has had several variations of better watermarks - the ones they just replaced with this one would do. The one before that was even more "in your face". They don't need any help to come up with something better and they chose this weak one with a full understanding of its limitations.

The way things are now - none of the stock licenses mandate a maximum size for online usage, for example (which they used to do) - a watermark is just there to alert honest people that they need to license the image. On the other hand, making it so easy to steal can get out of hand. Ask Hyundai and Kia about their settlement.

And I know, stock agency contributors don't have anyone to go to bat for them as work is stolen wholesale.

This is another voice actor situation - Greg Marston did voice work for IBM in 2005, but IBM used that to train an AI voice and then sold it to Revoicer. No compensation for Greg. He understood IBM could use that recording in perpetuity, but not for new uses, essentially competing with him in the here and now for voice work.

https://www.ft.com/content/07d75801-04fd-495c-9a68-310926221554

�[Marston] is working in the same marketplace, he is still selling his voice for a living, and he is now competing with himself,� said Mathilde Pavis, the artist�s lawyer who specialises in digital cloning technologies. �He had signed a document but there was no agreement for him to be cloned by an unforeseen technology 20 years later.�

https://www.linkedin.com/posts/gregmarston7_has-a-computer-company-cloned-this-mans-activity-7087772812093857793-uEg2/

(paywall)
https://www.nytimes.com/2023/09/23/opinion/ai-internet-lawsuit.html

"Tim Friedlander, founder and president of the National Association of Voice Actors, has called for A.I. companies to adopt ethical standards. He says that actors need three Cs: consent, control and compensation.

In fact, all of us need the three Cs. Whether we are professional actors or we just post pictures on social media, everyone should have the right to meaningful consent on whether we want our online lives fed into the giant A.I. machines.

And consent should not mean having to locate a bunch of hard-to-find opt-out buttons to click � which is where the industry is heading.

Compensation is harder to figure out, especially since most of the A.I. bots are primarily free services at the moment. But make no mistake, the A.I. industry is planning to and will make money from these systems, and when it does, there will be a reckoning with those whose works fueled the profits."

...The picture is aesthetically pleasing and most people will recognize it as a robin, but a lot of the details are wrong.

Here is how it should look like:

https://upload.wikimedia.org/wikipedia/commons/f/f3/Erithacus_rubecula_with_cocked_head.jpg
...

I'm not an ornithologist, but I've attached what Google shows for the American Robin and the European Robin (what you were showing). The AI robin isn't either of them, although as you say it's a pretty bird on a pretty branch.

That's (IMO) a type of visual pollution that AI has created - all sorts of things that are kind of like something you recognize, but not really. I think the difference matters when it comes to photo-realistic content. For fantasy creations, anything goes.

In time, the cynic in me says that many will argue that any differences between reality and genAI representations of it aren't important anyway. I've already seen some of Adobe Stock's AI images show up in Google searches for real places - and the Capo San Marco lighthouse looks nothing at all like the AI version of it.

For now, the pollution level is small enough to know which is which, but the volume of pretty-but-misleading images is growing rapidly.

When someone points a camera at something real, even if the person taking the image isn't an expert in the thing they're photographing, what is captured is something that actually exists. Bad keywording is obviously an issue, and that can be misleading too, but that can at least be fixed (should an agency decide it needs to do something about it). The rabbit hole of labeling AI "hallucinations" as real places, real animals, real birds, etc. can't be fixed.

Adobe's rule that specific place or property names shouldn't be used when labeling genAI images is a good one. It just isn't enforced. I think it should extend to species names for plants, birds, animals as well unless it's an illustration.

@Mantis, anecdotal, but very interesting.

My crystal ball is broken, but my guess is that two things will keep stock agencies in business (though volume might not grow, so investors won't find it "hot" any more).

1] Images where the specifics matter - you mentioned a robin, but I have a collection of genAI bird and animal images from Adobe Stock where they're not even close to the real thing; same with places, machinery, wheelchairs, stairs, ladders, etc.

2] The "Midjourney look" has become so ubiquitous that in a little while it will be overexposed and viewed as tired - been-there-done-that - and the trend will shift to something else. It's not about extra/missing limbs or digits, but I get an overwhelming sense that the photo-realistic people photos are from some vast clone factory and that you're seeing the same few people everywhere.

Back when microstock was a baby, some buyers were happy to see what they viewed as more realistic people and settings than the traditional stock images. That was because by and large the models were friends and family and not professionals. As microstock wanted to grow, it tried to look more like the the stuff they were replacing - and some stock photographers from the earlier generation came to the micros. Midjourney images are a refreshing thing, for a little while (if you ignore the origins of their data training; I'm guessing your company isn't worried about being sued over use of Midjourney images?).

And I have to add, for the bazillionth time, that I keep hearing how much better things are getting with the "oops" images, but that's not what I see looking at new approvals in Adobe Stock's 16.8+ million genAI collection.

If you wanted a saxophone player, neither of these would be "close enough". Nice lighting and engaging look, but the details are effed up. These were new this morning.

 

iStock does not allow AI images (although they let a few slip in). They are down to a very small number.

Someone knows how many days appears the money when it is withdrawed from Adobe Stock? I withdrowed on september 12th and not news about on my PayPal account. Thanks!

I normally request money at the end of each month, but this month I too requested a payout after the Firefly "bonus" showed up. I haven't yet been paid either. They say 7-10 days (I think - it might say 7-10 business days).

I can't request my end of the month payout until the earlier one is paid - you're not allowed to make a request if there's a payment pending - so I hope they sort this out next week.

British actor Stephen Fry found his voice had been used to narrate a documentary without his knowledge or participation. The article says they used his narration of audio books to get the needed samples:

https://www.theguardian.com/technology/2023/sep/20/it-could-have-me-read-porn-stephen-fry-shocked-by-ai-cloning-of-his-voice-in-documentary

I have to imagine there'll be litigation about this, but nothing so far.

https://nypost.com/2023/09/18/harry-potter-narrator-stephen-fry-says-ai-was-used-to-steal-his-voice/
https://variety.com/2023/film/news/stephen-fry-ai-stole-voice-harry-potter-audiobooks-1235727795/

This is about the music business and different ways to license sounds that would allow artists to retain the rights to the sound of their own voice

https://www.theverge.com/2023/9/21/23836337/music-generative-ai-voice-likeness-regulation

"AI poses a tricky problem for labels and musicians under copyright law: establishing ownership of a song that sounds like an artist�s overall output but does not feature a direct copy of any particular work. "

"It isn�t just AI-powered music platforms looking at licensing artists� voices; even established record labels believe this is a good start. The Financial Times reported Universal Music is in talks with Google to license artists� voices and melodies for generative AI projects."