No doubt about it – AI (artificial intelligence) slop is everywhere. You’ve probably seen it without even realizing you’ve seen it, the way it’s permeated advertising, social media, and even LinkedIn! Being able to detect it is getting even harder than avoiding it, with new advancements coming out every day. So much so that, while I’m going to try to offer helpful ways to detect AI slop, it might become outdated fairly quickly.
Why ‘Slop’?
The term first appeared online in 2022, when generative AI art began to be used frequently. Though other derogatory terms were also used, slop is the one that stuck, with 4Chan and YouTube being the sites where used most prevalently. It refers to the fact that generative AI art, when first introduced, was typically of very low quality: people had multiple hands or eyes, and other facial features were a little melted, among other poorly constructed details.
The term has now stuck and, for many people, applies to anything that is generated by AI – even if it is easy to mix up with real photography. So now, if something is AI, no matter its quality or number of fingers, you’ll likely see it referred to as "slop" because of widespread concerns about AI.
How to Detect AI Slop (for Now).
This section explores simple ways to tell if something was AI-generated. It’s important to remember that as technology rapidly advances, some of these ways of telling will become outdated. That said, it’s also good to have a critical eye on what you are taking in. For example, clicking on the profile where the questionable art came from: if the content is largely the same (for instance, just different animals jumping on trampolines), then there is reason to suspect everything on the account is AI-generated. Sometimes they even have it in their profile, letting you know.
Either way, these tools can still be applied even if the video or image was edited by a person after being generated by AI.
Background
Scanning a picture and noticing details are giveaways that an image or a video was generated. Most people only pay close attention to the foreground, so items in the background don’t get as much attention when the prompt is being written.
The image on the left, for instance, shows a war scene. But the sign on the shop is indecipherable. Closer inspection also shows details that are a little off, like some of the cars aren’t full-size and don’t even have wheel hubs. The fire itself is also a little odd, not contained at all, but it appears not to be spreading because of how many people are just standing near it or actually in it.
Things that don’t make sense, or details that are a little fluid when trying to make out what they are, are all indicators that the image is AI-generated.
Camera Position
For a time, videos of animals knocking over Uber Eats or other delivery items on porches were everywhere on social
media. These were all AI-generated videos and would be an example of how to tell if the account is AI. Accounts with only videos of animals knocking over Starbucks deliveries and other random items are good evidence of an AI account.
One way to tell if a video is AI that doesn’t require investigating is to notice where the “camera” is on the video. In the image to the right, the camera in question doesn’t seem to be attached to a wall. Notice how the other side of the porch is open to the elements and only has a railing? Generally, porches mirror each other when they’re small and relatively simple in design. Maybe there is a home with a wall there, but if you believe it doesn’t, it could be an indicator that the video is AI-generated.
Morphing
One feature that is a little harder to pick up on at first viewing is shifting. For instance, a cat’s tail might be sleek, and they move, and it’s fluffy, or someone’s arm becomes a bit longer than it was before. This is harder to spot because, obviously, cats’ tails get fuzzy when they're excited or upset. In the examples below, it’s fairly obvious: the pancakes the man is about to eat suddenly blur in the middle of the video. Sometimes, the morphing into something else is a bit subtler, with something in the background shifting while your attention is on the foreground, so it’s important to pay attention to the whole image rather than just where your attention is drawn.
Watermark
You don't see it as often anymore, but sometimes there's a watermark in the image indicating it was generated by AI software. These can be removed, and not all of them have it, but if you see a symbol like the one on the left, you know it’s AI.
Too Shiny
There is a certain texture to AI-generated art. It can be a little too shiny. The photo on the right of a burger might look appetizing at first glance, but its features are also a little too clean. This is generally because AI doesn’t understand lighting or shadows, so everything is just a smidge too perfect, until you notice the bun-sized pickle slice, the precisely ruffled lettuce, and the cherry on top.
When it comes to humans, the situation is much the same. One theory has it that AI was trained mostly on pictures of models, so every person it generates looks a bit too glossy and over-manicured.
Other Features
Beyond all of the things I’ve shown here, one way to tell that something might be AI-generated is to also pay attention to things that don’t seem quite right in the image. The obvious one is when they have an implausibility, like three legs or six fingers. As mentioned above, AI doesn’t quite grasp lighting, so lighting on the face not matching the lighting in the background is a big giveaway. Other little details to look for include backpacks that just merge into shirts, or buttons that don’t quite make sense.
The picture to the left of a person viewing himself in the mirror is one such example. He’s wearing a short-sleeved dark shirt, but in the mirror, he’s wearing a long-sleeved light blue shirt. Also, the angles of the head and face in the reflection do not match those of the person in front of the mirror.
Why Do People Choose to Use AI Art?
Some people do it as a sort of harmless scam, especially on social media. They can quickly "create" content to post. If that content gets people to interact with it, it will generate revenue, or the potential for revenue if they are building a following.
It’s cost-effective. Paying professional artists and photographers is not cheap. So, people choose to have an AI do it because, at the moment, it’s free to use.
Sometimes people just create images because they think it’s fun to see themselves in different styles. For a while, a lot of people were turning themselves into anime characters using generative AI.
Often, people do it for much more pragmatic reasons. They might need a quick image for a presentation, so they type up a prompt and get what they need in seconds. Instead of asking a coworker to take the time to create the art, they can get what they’re looking for and move on without another thought.
It must be said that the convenience, cost-effectiveness, and user-friendly packaging of generative AI hide serious issues. For one, AI requires huge amounts of energy, water, and space. Increasing demand for those limited resources typically means higher prices for everyone. The massive AI data centers have a significant environmental impact. That is not to mention the ethical and intellectual property issues that AI companies blithely disregard.
The speed at which AI is improving is both utterly predictable and deeply troubling. It feels to many people that we are already past the point of no return, where we can no longer trust any image as fact and the companies spearheading this revolution seem never to have taken a moment to consider what that will actually mean for the humans who will, one way or another, have to share the planet, at least for awhile, with their product. It's probably time we ask the question: are all the considerable downsides of generative AI really worth the LOLs from a picture of the Pope strolling around in a Balenciaga puffer jacket?
What do you think of AI-generated media? Let me know!
Related Content
HCPL Blog:
Is Artificial Intelligence a Modern-Day Frankenstein’s Monster?
Is AI OK? A Librarian’s Thoughts on the Ethical Use of Artificial Intelligence
Transhumanism in the Wake of Artificial Intelligence
Books:
Unfinished: The Role Of The Artist In The Age Of Artificial Intelligence
Add a comment to: What is AI Slop and How Can I Tell?