By: Jackson Liu
AI, short for artificial intelligence, was first made by researchers in 1956 at Dartmouth College. It was very primitive compared to AI today, with much weaker computing and thinking skills. AI is an algorithm that can learn on its own and can partially simulate human emotions. Since then, as researchers continued to perfect AI intellegence, it has been getting smarter up to the point where it can even generate photos or whole articles in seconds. Social media users have also started warming up to the idea of AI generated photos. Now, more and more photos generated by AI are appearing online. Here is how to identify which photos are AI generated and which are real.
One way to identify AI is to focus on little details. AI makes photos by analyzing real photos, but often makes mistakes on small details such as hands and text. AI is terrible at drawing hands. AI drawn hands are often distorted or have more fingers than a normal hand. This is because those small things don’t appear that often in pictures and is very complicated. It only occasionally draws hands perfectly, in which case you could look at other features like the background and the face.
AI image generators often make mistakes on text in the background. Most of the time, when you look at AI generated photos, the text inside is gibberish. For example, if you make it generate a photo of a person in the street, logos and other text would look like scribbles. A regular Mc.Donalds logo would look like a strand of slime tangled around.
AI also sometimes makes mistakes when drawing a human face. Normally, if a human is the centerpiece of an art piece AI would have no problem making it look human. But often overlooked are humans in the background. They often lack facial features, such as eyes and noses, and look like misshapen mounds of sludge. There is a rare case when AI draws the faces in the background accurately, though then it might be too perfect.
Another way to identify AI is to look at the quality. AI tends to make everything perfect, like for example when they draw faces they never draw imperfections unless specifically instructed to do so. They also tend to draw things in a perfect, fantasy background where nothing is ugly and everything is perfect. A little too perfect. There’s sometimes a rare case where AI draws everything perfectly and you can’t find a single problem. In that case, you could do a reverse image search, where you find the same picture on other parts of the internet. AI generated photos appear less than normal photos. A reverse image search could also lead you to the source, where you can identify its legitimacy through hashtags and comments.
In conclusion, AI is advancing rapidly but still makes many mistakes when it comes to image generation. Next time you see an image on social media you’ll know if it’s AI generated or not. Soon, though, in another few years AI would progress so much that it’s completely impossible to identify real and fake photos.