I read an article about a new AI (artificial intelligence) image generation program called DALL-E that's supposed to be able to generate life like images based on a sentence that you give it. I was all excited thinking that it would be a great way to create some placeholder photos for Christmas cards and the like without needing to worry about model release forms and all that.
Well... er... um... maybe not! Seriously, I can NOT stop laughing!
But maybe it would be useful for something else? Anyhow, here's the URL if anybody wants to try it out. https://labs.openai.com/
The US ruling mentioned somewhere above was that images that are straight out of an AI program cannot be copyrighted because they weren't created by humans. (I have issue with that premise but that's beyond the scope of this thread.) However, AI generated images that have been edited/modified CAN be copyrighted. There is no standard yet as to how much modification is required to show meaningful human input so that would be subjective according to whoever at the copyright office is reviewing your application.
We all know that when you create an original work, its yours by default. For infringements found on the web you can initiate a DMCA take-down notice but those work on the fear-factor. They are scary and people usually don't want any hassle. But the entity receiving the notice can file a counter-notice. If that happens the claimant has 14 days to file a lawsuit, or drop the claim. And there's the rub: Infringement cases are the jurisdiction of the Federal court and you need to have an officially registered copyright to file a suit. It's $35-$55 per work, lots of red tape, and months of waiting to get an official copyright registration. I'm willing to bet that outside of the licensed brands, the overwhelming majority of work found on PODs and stock sites has not been officially copyrighted. For the millions/billions images scraped from the web to train AI, one would have to be able to pinpoint a specific work that was reproduced in a generative content result AND hold a registered copyright for thaat work, to sue for infringement.
I write this because discussion of AI art always includes lots of mention of lawsuits, but without officially registered copyrights, those suits aren't going to happen. The situation we have now basically runs on the honor system and the fear factor, it's all bark and no bite unless you've an officially registered copyright.
And FYI did you know that for images generated solely by AI, nope, you can't copyright them, but you can take a photo of that work and copyright that because photos you take are created by YOU? Interesting, no?
At issue is whether the act of training violates copyright. The output images themselves don't violate copyright (per the judge) since they are not similar enough to any original input images to infringe. Whether or not training constitutes fair use is an unprecedented question and may wind up in the Supreme Court.
I agree that that is THE question and whether we agree or not with the outcome we'll all have accept whatever is eventually decided in the courts. My personal layman's opinion is that it is fair use. I could download every image on Zazzle and as long as I was just using them for my own research & development purposes and not re-creating and re-publishing them, there's no legal issue.
It's very hard for me to put into concise words, but I think the heart of the issues with AI lies with the users and not the technology behind it. As with anything in life there will always be people who will abuse the system and thus ruin it for everybody.
This forum automatically saves posts-in-progress at regular intervals. In case such happens again, you can go to your forum profile page and look down on the right for the section that has your Drafts in it.
I don't know why we keep referring to my alternative medicine symbol because that's just commercial stuff.
It was just an example. Your motivations for creating it and intended use & meaning of it are totally irrelevant. The point was that you had to call on your pre-existing knowledge of what a caduceus looks like to create it. You didn't create it, then go online and be like "Oh, wow, what a crazy coincidence, my art I just made from my own imagination happens to look just like this common symbol that's been around for thousands of years!"
Noone's doubting it's your own original work that you didn't copy another's idea for and that's also the point. You couldn't go back and pinpoint any one specific image out of all the ones you've scraped in your lifetime that you infringed on to make yours. Thus with the millions/billions of images an AI program (not a "robot") was trained on. It's very unlikely one could look at a generated output and identify any one specific image from that knowledge base that's being infringed on.
Here's an old image I generated last year with the simple prompt of "bouquet of wildflowers in a modern green vase"
Here's one I just now generated with same prompt:
They don't look anything alike. Same as if you, me and Connie were given this prompt to draw ourselves. We'd all have to draw on our pre-existing knowledge base for vase, bouquet, green, wildflower .... but our resulting works would be wildly different. More to the point, you can do an image search for "bouquet of wildflowers in a modern green vase" and not find anything that looks remotely like either of these images. Likewise if you use these images to do a reverse image search. There may be some similar elements or similar compositions, but nothing remotely close enough to the whole to say hey, this is infringing on this one image of a vase that was included in the millions the AI was trained on to learn what a vase generally is...
This is over-simplifying the situation, I know. There are exceptions and that leads to where I said for me the problem is the users of it, not the technology itself. All this seems to be moot now though as as KeegansCreations pointed out, the main issue at legal hand now is not whether AI is reproducing copyrighted images in generated works, but rather, was scraping images for training without explicit consent a violation of anyones rights?
First of all, please stop writing "you infringed". I have a legal background and that is slander! I've taken screenshots btw. Would you like if people go through your shop and make this rubbish personally attacking you and your work to prove some point? Move on. You can use general examples without making it personal, can't you?
The only one making things personal is you. I haven't slandered you, I haven't attacked you or any or your work, I haven't said you've infringed on anything. I actually have been saying exactly the opposite to make my point. I said your image is very nice (because it really is) and used it purely as an example of how humans use their pre-existing knowledge base of what things look like to create their own unique works that aren't infringing on anything. And that came about in response to your comment that AI just relies on images it scraped from the web and I commented it's the same for humans and ... well here we are with you all offended because you don't understand the argument and me risking the wrath of the Mods for continuing to respond to your willful ignorance.
Good day.
