image credit – finance.yahoo.com
If you want to know if a picture was made by AI, look for things that don’t make sense, like earrings that don’t match and faces that don’t look right. Always check the text and hashtags in the image’s description and title to see if they mention AI software. You can use reverse picture lookups and GAN detection tools if nothing else works.
An AI-generated photograph is any picture that was made or changed using so-called “artificial intelligence” (AI) software built on machine learning to add or change synthetic content. As AI picture generators like DALL-E 2, Midjourney, and Stable Diffusion get better at making realistic images, some people have tried to make fake photographs.
It’s a big problem if someone uses an AI image of someone without their permission and shows them in a way that makes them look bad on purpose. How can we tell if a picture is real or if it was made by AI?
How to Find Out if a Picture Was Made by AI?
At the level AI is at now, it’s generally easy to tell if an image was made by AI or not just by looking at it. Fingers or glasses don’t match up. Sources of light don’t make sense. The backgrounds are bent.
But if you’re still not sure, there are other, more technical ways to look into a picture. We’ll talk about that later. First, let’s talk about the most obvious tip on the list.
Check the comments and captions for the photos
People who play around with AI and post the results on social media like Instagram will often tell you right away that the picture is fake. If it’s not clear that the picture is fake at first glance, look at the text.
Usually, it will say something like, “This picture was made by feeding my photos into AI” or “This image isn’t real. It was made using (Name of Application). For e.g. Midjourney.” They’ll write that it was made by Midjourney and include hashtags like #aiaart, #midjourney, #mjv5 (for Midjourney version 5), and so on.
You can also get a tip from the comments. The original poster might admit that the picture was made by a computer, but if they don’t, sharp-eyed commenters will notice and point it out.
Check for Oddness
Even though AI is getting better quickly, the pictures it makes can still look scary. At the very least, the uncanny valley effect shows up in many of AI’s photorealism efforts. Look for anything that seems out of place in the picture, like:
- Hands and teeth that look strange
- Faces that are crooked or not symmetrical
- Mismatched jewellery, such as earrings
- Text that was hard to read in the background. Watermarks on pictures used to train the program that were smudged.
- Textures that are too smooth
Watch out for things like eyeglasses that might blend in with the skin of an AI-generated person.
Use a GAN Detector
Generative Adversarial Networks, or GANs, are used to make images by AI. You can add an extra layer of security to your attempts to verify a picture by using apps and plugins that try to spot fake images. For example, if you right-click on a personal picture in Chrome, a plug-in will check to see if it was made with GAN.
Results from these programs are hit or miss, so it’s best to use them along with other methods and not count on them completely. Just like AI image generators, this technology will improve, so don’t dismiss it.
Do a web or image search
There’s always the trusty Google image search if you’re unsure of what you’re viewing. These days, you can simply right-click an image, and Google will return images that are visually similar.
To attempt and determine where the image originated, you can also utilize the “find image source” button at the top of the image search sidebar. The image you are viewing might not be of a genuine person if the search produces no results.
Find other reporting on the same event if the photograph is used in a news story that might be a misinformation piece. It could be a hoax if no other media cover it, particularly if the incident is dramatic.