International Business News – Text-generated images are a new technology. Some argue that a recent proposed rule in the UK is detrimental to copyrighted photographs.
Matt Growcoot, news editor for the professional photography equipment review site PetaPixel, writes, “For photographers, they need to ask themselves. For photographers, they need to ask themselves a few questions: How would you feel if you saw an artificially intelligent image based on your photo? The second question is, what are you going to do about it?”
“In the age of the Internet, no other type of creator’s rights are trampled on as often as the photographer’s. Artificial intelligence image generators have the potential to be the next trampler.”
If you enter an idea into a text image generator like DALL-E 2, for example, you want a picture of the legendary photographer Ansel Adams. Ansel Adams style, it will provide you with photos that look like the ones Adams took. How does the text image generator know what these photos look like? Obviously, the AI is trained based on these images, which means that the images are fed into the system. However, these images are protected by copyright, so why can they be used for AI training?
OpenAI, an American AI research company, cleverly sidesteps this question and says it has trained DALL-E 2 based on “publicly available” images, including, of course, those that are copyrighted. Just because images can be found in a Google search doesn’t mean they are free to use. This is a gray area of the law, but U.K. government officials are interested in clarifying the issue.
A new proposed exception in U.K. copyright law makes it clear that it is legal to make any online content available to AI. The new regulation would allow machine learning programs to freely use all images posted online, according to a scathing review of the proposed change by the Association of Photographers (AOP).
AOP writes: “On a practical level, this means that the AI bot/crawler will scan or read any digital images on your website or social media accounts and extract any data that the bot has been programmed to search for (images and metadata embedded in the original source and all versions available elsewhere).”
“The bot will make copies for the AI platform so that the AI can learn from them and create new images.”
It’s hard to argue that companies like OpenAI aren’t already doing this. But the U.K. wants to make such practices a safe and protected legal practice. It’s not hard to infer that the U.K. envisions itself as a haven for AI companies and is willing to sacrifice photographers’ copyrights to make that happen, industry insiders say.