At the time of writing this post, Ai images are hyper-realistic and are quite difficult to distinguish from real images. However, thankfully there are still small, concrete, distinguishable, differences that can be found between real photographs and Artificial Intelligence generated photos. Ai does seem to be improving its quality but there are certain kinks that have not been worked out. And those are the small details that will protect you from misinformation.
There are now countless free Ai image generators. Large tech companies are considering adding watermarks or small digital footprints to AI generated images in an effort to help viewers depict which images are real or fake. But currently, the only person who can protect you from misinformation is You. After this article, you will be able to tell with 99.9% accuracy which photographs are real or fake. Lets dive in to the specifics.
Dangers of Ai Images
One of the main dangers of Ai is the fact that it can create very believable looking images that can then be used for harm. This has become such a huge problem due to the ease of use of free ai image generators and Character Ai. Currently there are major problems in schools with children using Ai to generate hyper-realistic images of classmates and teachers without clothes on. This same problem is also happening with celebrities and there are currently no repercussions or consequences for it. Ai has already been used to generate fake images and videos of political leaders to deceive the public. For this reason, you NEED to know the difference between these images and real images. In fact, here are some photographs you might recognize :
Every single one of these photographs was generated by Artificial Intelligence and still went viral enough to make it onto main stream news and media. These three images are viral images of the Pope, Donald Trump, and Sofia an AI creation.
The image of the Pope actually made it onto multiple news channels on live tv and TMZ. The image of Donald trump being arrested recently went viral across the internet with hundreds of thousands of negative comments about him resisting arrest. Of course that is not true, he did not resist nor is that image real.
The third image is a picture of a girl named Sofia. Sofia is a social media model with millions of followers. But Sofia is entirely created by Artificial Intelligence. She is not a real person she is entirely made up by Ai. Yet she has millions of followers who still believe she is real !
Read on to find out how you can tell that these images are fake.
How to tell the difference between real pictures and AI
Step 1: Anatomy
The first thing you should do when you see a questionable image on the media is pick a small portion of it and look at the details. Which portion of the image should you pick? If there are people in the image, then look at their body and count their fingers. Artificial intelligence, for some techincally advanced reason, tends to mess up the human anatomy. This is especially true in pictures that depict motion or when a hand is not a main detail in the image.
Scroll up and take a look at the picture of Donald trump being arrested. Pay attention his legs. Notice anything? Trump’s legs, as well as the officer on the right, just don’t make any sense. It looks like their legs are broken and tied in a knot. And if you look closer you’ll see the officer on the right doesn’t even have legs. This is the most common tell of an Ai generated image.
Here is an example of an AI image with the wrong amount of fingers:
At first sight, this image looks very real. But take a look at the hand on the right side of the image. From left to right: there is a pinky finger, then a ring finger, then what appears to be an extra finger, then a middle finger, then an index finger. That extra finger is a tell that this image is generated by Ai. See how small the difference can be?
Did you notice any other tells that this image is fake? Another tell is the awkward angle of the finger he is using on the trackpad of his laptop. And yet another tell is that there seems to be no space bar on his keyboard. One more tell is that the pocket on his suit jacket is sewn at a weird diagonal angle. And finally, the left half of the image shows that the table is round, but the part of the table where the gentlemen is sitting has a flat side. So that is either a very awkwardly shaped table, or the more likely reason is that the image is Artificially generated.
Read on
Step 2: Small Incorrect Details
It is rare that an image passes the first step, but if it does, then the second step is to look for small incorrect details. These details usually require you to use some common sense and knowledge to figure out. Kind of like one of those old internet games where you search for little objects in the image.
In the case of fake images, these differences can vary widely. Some tend to be small words or signs in the background of an image that have random symbols on it rather than an actual language. Others tend to be mistakes in clothing like a zipper or jacket opening that doesn’t fit correctly onto the jacket..
Now scroll up and take a look at the picture of Pope Francis on the right. Did you notice his jacket-opening trails off and ends at the top of the crucifix he is wearing? That is an example of using common sense. Did you also notice his crucifix? This is where knowledge comes in to play. A crucifix should depict the body of Jesus Christ nailed on a cross with a small inscription above his head. The crucifix in the image looks like the Artificial Intelligence tried to depict that but was lazy with the details and the result is that it looks like a smudge of metal on the cross. Also, the chain holding the cross around his neck is depicted on the left side of the image but vanishes once it touches the crucifix.
Step 3: Texture and Depth of Field
I debated putting texture in with Step 1 but it became apparent to me that texture is not something that is inherently obvious to everyone. And if the first thing you do is judge a picture by its texture, then you might accidentally dismiss real pictures as fake. This is because photographers often try to soften the “noise” in their photos by softening the texture of the picture. So make sure texture is the last thing you look for in an image because it is not always obvious nor is it as reliable as steps 1 and 2.
This step requires a basic knowledge of photography to fully understand. Scroll up one last time and take a look at the image on the right, Sofia the artificially generated model. This image seems to pass Steps 1 and 2 at first sight. But the texture and depth of field of the photo is all out of wack. The texture is so soft that she appears to not have any smiling wrinkles nor any wrinkles in the crease of her elbows. Also, the texture of the tall grass behind her is so exaggeratedly soft that it almost looks like something out of an animated Disney movie. This of course can be done to a real image with photography and editing. That is why texture is more of an intuition thing and not solely reliable to someone who is not an expert.
Now here comes the last and most advanced form of finding an Ai image with your bare eyes. Depth of Field, on the most basic level is the pattern at which one part of an image is in focus and the other parts are intentionally blurred. I know that is not 100% accurate but that is the best way to describe it to someone who has never heard that term before.
When looking at the image of Sofia, the front of her body is the subject and it is in focus. Slowly as the distance gets further, it should get blurrier and blurrier. Is blurrier even a word? Maybe. This is why the plants on the left are kind of in focus and then as the distance grows they get blurrier. Well at that pace, when you look at the big green trees in the background, they should be much more blurry. They should look like a big green blob but instead you can still distinguish each tree apart and you can even see the tree trunks of each tree. A real image at that f stop (photography term) would not be able to depict each individual tree apart like that. Therefore the depth of field is simply wrong.
PSA
If you found this Post informative, helpful, or enjoyable, please share it with a friend or family member. This is such an important skill to have in today’s world. Thank you !