Friday, April 07, 2023
Watermarking ChatGPT, DALL-E and others could help protect against fraud and misinformation
Salon.com: Shortly after rumors leaked of former President Donald Trump's impending indictment, images purporting to show his arrest appeared online. These images looked like news photos, but they were fake. They were created by a generative artificial intelligence system.
Subscribe to:
Post Comments (Atom)
3 comments:
I remember it having this conversation when deep fakes were starting to take over the Internet. What do you do when you can’t tell if a photo is real or not? I’ve heard that there are professionals who have always been able to tell the difference between real and fake photos, however, nowadays, they cannot anymore due to the high quality of ai photo generators. For artists sake, I hope they do watermark is an eyes, so that way the artist integrity of their work will remain clear to the general public. I also think this would be helpful in terms of what can and cannot be used in a court room. Evidence is so often already fact that the advanced technologies available to everyone for a low-cost could cause lots of legal issues if not handled properly soon. However, this would mean it is nearly impossible to use AI are in theater, without breaking the illusion of the show, because the watermark would be recognizable and likely alien within the world of the show.
Posted by B Hanser
The level of realism that generative AI can now achieve is getting kind of scary. I saw some of the AI generated photos and videos of Donald Trump getting arrested, and while most of them were clearly not real, some of them were hard to tell whether or not they were. I’ve heard a lot of AI generated audio clips of presidents and other famous people talking recently, too, and it’s incredibly difficult (at least for me) to determine if it’s fake or not. Thankfully, the audio is usually about something stupid like Donald Trump and Joe Biden in a drive-thru, but it’s scary to know that people could make stuff like this if they tried. The development of AI is simultaneously thrilling and terrifying. There have been so many articles about legal problems with generative AI, and I think that while a watermark wouldn’t solve all of them, it’s definitely a step in the right direction.
Post a Comment