“Photoshopping” images is the act of altering an original image to the extent that it conveys something else entirely. It’s harmless use results in funny images and memes but it’s regularly used as a tool for propaganda and spreading misinformation as well. Such images can often trick social media and even news outlets to the point that they end up going viral. Adobe wants to reduce the role that its software can play in the spread of misinformation so it’s going to use artificial intelligence to catch photoshopped images.
Adobe says that using artificial intelligence to spot fake images will go a long way to “increase trust and authenticity in digital media.” Digital modifications to photos leave behind a trace in most cases and these traces can help the AI determine whether they have been changed in any way.
Forensic tools have long been used to find these traces by looking at the noise distribution, inconsistencies at the pixel level, and lighting. However, Adobe believes that artificial intelligence can do a better job while being faster and more reliable.
The company’s AI will look for three major signs of modification – splicing, cloning, and removal. Adobe and a team of researchers from the University of Maryland have trained the AI on thousands of fake images so that it understands what it has to look for. Adobe says that the AI can now achieve the intended result in seconds.
It does mention that the software isn’t perfect and that there will always be room for improvement but it does “provide more possibility and more options for managing the impact of digital manipulation, and they potentially answer questions of authenticity more effectively.”