The British government is considering a new law to regulate AI generated images, making their origins clear to all.
On the surface, new regulation being considered by the British government to require the labelling of any image created with AI seems a good idea. After all, AI is reaching a point, and in some cases has reached a point where artificially generated images are indistinguishable from real photos.
However, we all know how well intentioned, but poorly thought out laws can cause more problems than they solve. No specifics for the proposed regulation have been released, but the realm of AI image generation is fraught with grey areas.
For example, would a square photograph that has been changed to a 16:9 ratio with AI generated content to fill the sides be considered to be an AI generated image, and therefore required a label? Or, how about a photograph that has had power lines removed using an AI process? Furthermore, would the new regulation apply to artwork? After all, most of the fears from AI generated image content comes from the risk of fake imagery being used for nefarious purposes. Could the same be said of imagery that looks like comic artwork, or abstract?
Enforcement of such regulation could be pretty much impossible unless it was taken up unanimously on a global scale, and indeed this is what the Prime Minister hopes will be the case. A British AI safety agency is also in the process of being set up to monitor the various AI companies to ensure that they keep to their original intentions. Meanwhile, Google has also pledged to label AI generated content to inform users about an image's origins.
This is all being done with the best intentions, although given how fast AI is progressing, as well as factoring in the more criminal aspects of society, not to mention hostile governments around the world, it's hard not to see such discussions as bringing a water pistol to an assault rifle fight.
Regulation works when it is easily enforceable, but unfortunately, the internet works by relying on an enormous amount of good will and community. Unfortunately, as we've seen all too often, the internet is in reality a Wild West. Then there's the question of what the punishment would be for not labelling an image? Not to mention the question of who would be liable? If an AI image is posted anonymously, and it's so good that it is indistinguishable from reality, and it spreads around the internet, who bears the legal responsibility?
The legal aspect of AI is something that we will, ultimately, have to get to grips with, but I fear it's going to be a very long, drawn out process, with many speed bumps along the way. There's no question that regulation will be needed, but the elephant in the room is how it can ever be done with any degree of true effectiveness?