SoylentNews is people
SoylentNews
from the artificial-artificial-intelligence dept.
On Monday, Adobe announced that its stock photography service, Adobe Stock, would begin allowing artists to submit AI-generated imagery for sale, Axios reports. The move comes during Adobe's embrace of image synthesis and also during industry-wide efforts to deal with the rapidly growing field of AI artwork in the stock art business, including earlier announcements from Shutterstock and Getty Images.
Submitting AI-generated imagery to Adobe Stock comes with a few restrictions. The artist must own (or have the rights to use) the image, AI-synthesized artwork must be submitted as an illustration (even if photorealistic), and it must be labeled with "Generative AI" in the title.
Further, each AI artwork must adhere to Adobe's new Generative AI Content Guidelines, which require the artist to include a model release for any real person depicted realistically in the artwork. Artworks that incorporate illustrations of people or fictional brands, characters, or properties require a property release that attests the artist owns all necessary rights to license the content to Adobe Stock.
[...]
AI-generated artwork has proven ethically problematic among artists. Some criticized the ability of image synthesis models to reproduce artwork in the styles of living artists, especially since the AI models gained that ability from unauthorized scrapes of websites.
(Score: 2) by Sjolfr on Saturday December 10 2022, @03:48AM
... the movement for AI rights will begin shortly.
(Score: 0) by Anonymous Coward on Saturday December 10 2022, @04:04AM (11 children)
Aren't these AI images in the public domain?
(Score: 0) by Anonymous Coward on Saturday December 10 2022, @06:49AM (10 children)
Especially if they are trying to profit from this.
(Score: 3, Interesting) by sjames on Saturday December 10 2022, @08:37AM (9 children)
It's highly questionable that any sort of license would even be needed other than permission to view. AI image generators train a neural net on many images (with descriptive text encoded in the inputs) in order to do what they do. They do not 'sample' like in some forms of music. The generated images contain no pixels from the training set. If I don't need a licence beyond a license to view (implied if it's on the web) a bunch of images to train myself to paint/draw/photograph better, why would the AI?
(Score: 0) by Anonymous Coward on Saturday December 10 2022, @09:34AM (8 children)
People have been successfully sued for derivative works.
Those people didn't need a license to view copyrighted work but they sure needed to pay to produce derivative works.
(Score: 2) by bzipitidoo on Saturday December 10 2022, @04:04PM (1 child)
>they sure needed pay to produce derivative works.
Why?
Such a requirement just gums up the works. When did you have in mind that payment should be made? Likely, you thought of it as a "sale", to be paid at the point of sale? If so, it puts a lot of risk and burden on the buyers. They don't know if their expenses will ever be recovered. If there is to be any payment at all, maybe it should be a cut of the profits, if any of those are made?
(Score: 0) by Anonymous Coward on Sunday December 11 2022, @04:44PM
I used past tense for a reason. Go ask the judges why or read the judgement: https://en.wikipedia.org/wiki/Rogers_v._Koons [wikipedia.org]
(Score: 0) by Anonymous Coward on Saturday December 10 2022, @05:01PM
Fonts are a bit weird when it comes to copyright. As usual, rules vary between countries. In the US, the actual appearance of a typeface is not eligible for copyright protection so in the US such images will not be any problem at all.
What is copyrightable are the font files since Type 1, TrueType, etc. font files are actually computer programs (and software is subject to copyright). But once text is rasterized/printed/whatever, the resulting images are not.
(Score: 2) by sjames on Sunday December 11 2022, @01:25AM (4 children)
What if it does? As long as they're not copies. Are you saying there can be one and only one painting of a cowboy riding an ostrich in the entire world?
(Score: 0) by Anonymous Coward on Sunday December 11 2022, @04:42PM (3 children)
https://en.wikipedia.org/wiki/Rogers_v._Koons [wikipedia.org]
(Score: 2) by sjames on Sunday December 11 2022, @11:37PM (2 children)
That's not quite the same issue. Koons made an actual intentional copy of Rogers' specific photo. Had he just taken a different picture of a man and woman holding puppies, there wouldn't have been an issue. Had the sculpture been meant as a parody of that picture in particular, his parody defense would have held up.
(Score: 0) by Anonymous Coward on Tuesday December 13 2022, @05:50AM (1 child)
(Score: 2) by sjames on Thursday December 15 2022, @08:57AM
So I guess you're all out of ammo? If you can site something that might show there can be only one picture of a cowboy riding an ostrich in the whole wide world, please present it. If you can present something where there was an award for something not a deliberate copy of the original work, please do present it.
(Score: 2) by oumuamua on Saturday December 10 2022, @05:43PM (1 child)
Here we have AI producing art competitive with real Artists.
Already in the works: AI producing music and videos.
With ChatGPT we have AI producing flawless article compositions and basic code competitive with entry level programmers: https://www.youtube.com/watch?v=0A8ljAkdFtg [youtube.com]
AIs now beat humans in every known game.
Expect full self driving soon.
The AI naysayers have been proven wrong. Especially considering the models will keep improving every year.
Now something has to give, the old advice for those out of work: 'learn to code' no longer applies as the AIs will soon be coding.
Time to support UBI.
(Score: 2) by takyon on Sunday December 11 2022, @12:19AM
I'm not convinced by ChatGPT code yet but someone will probably turn Stack Overflow into a coding monster soon. I was impressed by the ChatGPT poems that I saw, and I would like to see more of them and try it myself.
The AI art is best with a certain amount of human intervention (i.e. work/labor), particularly a sketch or something for image-to-image translation, and fine-tuning with tools like DreamBooth. But it is amazing how you can get something you like (not necessarily 100% what you were going for) in very little time with so few tries, even using Stable Diffusion v1.5. Stable Diffusion v2 has taken steps to limit the artists included in the model and the ability to generate NSFW images [archive.ph], so there will have to be a great forking event soon. The depth2img and inpainting changes [stability.ai] look great.
You can get great AI art in a few seconds, but spending some time on it helps. Graphic designers will continue to exist, but maybe less of them. Artists being paid by commission will take big or small hits depending on their area. Their starvation might happen slowly enough that they don't notice until 25-50% of the work is gone. I don't think Greg Rutkowski [artstation.com] will starve, but maybe artists that aspire to be Greg Rutkowski will just give up before reaching that level and adopt the AI tools instead.
Music/voice look very hard. I thought voice would have been completely conquered a couple years ago, but the latest I've heard still sounded janky to me. Maybe the best code is proprietary and will be "responsibly" milked for cash for a while, or you'll need to do something like "voice-to-voice" translation to make your deepfaked politician sound convincing. I hope we see one of those markup languages for text-to-speech catch on, and the reverse, speech-to-text+markup. Once this is nailed down, it eliminates low-level Mechanical Turk style transcription jobs which I believe there are a lot of. Obviously, improving language translation is another big fish, and we have a subtle way of knowing when ML translations accessible to the public have reached stupendously high levels of quality.
Video may be easier than (it) sounds. We see that AI can already be used to create 3D models. What if it were to arrange 3D models and motion vectors in one step, and then convert each frame to something photorealistic while preserving temporal stability. Things off-screen would be factored in and you can have fully-raytraced lighting (which can be an imitation rather than classic ray tracing).
Games like Chess and Go don't matter. Get back in the cobalt mines, fucker! I think AI will have some good implications for video games, particularly open world games. It could eliminate jobs, or you could make the game world 10x larger with the same amount of man hours. We'll have to see how NPC AI can be improved without introducing cascading problems that break largely linear storylines.
Regulators, market forces, and public apprehension will suppress self-driving cars. It has to work so damn well, perfect 99.9% of the time doesn't cut it. If you want to replace big rig truck drivers, it has to be nearly flawless since those vehicles have a lot of killing potential and crashes can lead to multi-million dollar settlements.
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]