What can AI do?
Do you know for certain that the images you see online are real? Studies have shown that you probably can’t. For example, take a look at the following image:
The image on the left is the photoshopped image, but it is virtually undetectable to the human eye.
According to a study done by Adobe, their new software can detect photoshopped images with a 99% success rate compared to a 53% success rate by the human eye. That 53% success rate is no better than an average guess!
The AI is able to not only identify if an image is photoshopped but also where the photoshop happened. It is able to precisely show where the image was altered with a “heat map” of sorts. There are parts of a photoshop image that if zoomed even people can see it was unnatural. The AI can instantly and without (almost any) mistakes pick it up and mark the edits. This allows people to quickly see if an image was photoshopped and where.
What Impact Could AI That Tracks Edited Photos Have in the Future?
AI that can track photoshopped images are not yet mainstream and there is still work being done on it. But as technology increases and companies buy software or use open source software with its capability, those companies could implement the AI into their websites. This would hypothetically allow companies such as Facebook, Twitter, and other social media companies to automatically tag photos that are photoshopped. Realistically many of those large companies would not do that, but there would be websites or attachments online that could quickly allow people to verify if an image is photoshopped.
Because of this, people who edit images, videos, and audio could quickly and without much effort be recognized. This will potentially remove people who would spread fake news or create unrealistic body standards.
What Are the Potential Social Outcomes of AI?
You might be thinking, “Why does identifying slight changes using photoshop even matter?”. This matters for a couple of reasons.
Social media users will be able to trust images they see online- As a social media user, you probably run across photos of people every day whether it be on Instagram, Snapchat, Facebook, Twitter, etc. If all of these images show altered photos, how are you supposed to trust anything online? This will give transparency online and keep people from being deceived.
People’s worth and self-image would improve- By identifying photoshopped images, AI can help show people that everyone is human. All of the aspects that people try to photoshop out are a part of being human. By identifying images as photoshopped, people will be able to see that “Perfect” people don’t exist.
Will build a smarter community of online users- By identifying altered photos, audios, or videos, AI will help make online users less gullible to fake content. This will help stop crazy rumors and change people’s ideas of what a perfect person looks like.