AI Images and You: A guide to learning and understanding

Do you feel overwhelmed by AI?

Well, you’re not alone. In fact, I think it’s fair to say that, for most people, unless you are utterly tech savvy, you are probably overwhelmed by AI and the whole concept of it. There have been a lot of changes in what feels like a short period of time. I don’t think we can explain, solve, and understand all of AI in one essay, so let’s narrow our focus. Let’s talk about AI photo manipulation? That’s easier to understand, right?

Oh It’s not? Oh I see.

That’s okay, you’ve come to the right place. This is a space for those who are unsure of the what and why of AI photos and the how to of telling what is and isn’t AI. I am confident that by the end of this essay, I can have you walking out of here confident and comfortable in your newfound knowledge of AI photo manipulation and being able to tell what is and isn’t AI.

Let’s begin.

Defining AI Photo Misinformation.

Before we can begin to defend ourselves against AI photo manipulation, we must first learn what exactly it is that we need to defend ourselves against in the first place. AI has reached a point where, through simple prompts and questions, it can create images. CHATGPT, one of the biggest players in the AI market, is openly bragging about how advanced its AI can generate and create images. You can ask it to do images in certain styles , highlight certain objects, add and take away certain items. It really feels like unlimited possibilities at this point. 

Unfortunately, this is where the issues come in. Regardless of how you feel about creating art though AI and the morals behind all of that. The conversation can’t end there. We are not having this discussion to talk about the morals of AI art, no, this discussion is about what people do with AI created images and how to better spot them so that you don’t fall for their tricks. 

We understand why we are here. We understand what we are up against, now let’s tackle the real issue at hand.

Examples of AI art.

Let’s look at some images and play “AI art or not”

What did you think? Easy or hard? Did you do better or worse than you expected?

I think regardless of your results, it’s fair to assume that, even if predictable at times, AI is still getting better at image generation and the way AI as a whole is trending, this is a skill it’s only going to improve upon.  If you find yourself in the position of someone who is struggling at these “AI or not” tests and worry that you are in the minority of this, please do not. You only need to take a look at comments on TikTok videos to know that this is something that even “online” people are struggling to tell apart. This is something that I believe more people than are willing to admit are struggling with.

This is why we are here today, to learn about AI image manipulation and get better at detecting it. We have spent enough time talking about the how and why of AI imagery, let’s get in the real details of how to get better at spotting AI images.

So the first tip I want to point out that I think gives away bad AI imagery is text. If you are unsure about an image, examine the text, and you can usually spot some warped, weird text that gives it away.

Here are two great examples from legitimately huge companies.

WWE:

MLB:

You can see the way “Minnesota” is on the WWE poster or the writing on the team jerseys in the MLB ad that clearly, this is not normal imagery and that something is wrong with the text. This is clear AI usage by two giants of their respective corners of the earth.

Body parts! This is a huge one that, when you get comfortable with noticing, you will start to see everywhere, even before AI, back in the Photoshop days, bad or cut off body parts were a big tell of image manipulation. 

I think arms are a big one in general, they seem to get cut and shaped very weirdly. Here are some great examples of just bad AI imaging on hands.

AI images in general seem to have a lot of trouble with just little body parts like these, and paying attention to fine details like that in images can be a very easy tell when you are unsure or not.

Something broader to look at and pay attention to is just the overall background of images in question. Missing items, other body parts stretched and twisted, just the overall way the image is stylized being a little too perfect. The images almost have a shine to them that you would not get in a regular picture. These may not be as easy to spot when first learning, but these become clearer the more you become comfortable with the difference between AI and non AI images. 

A final point is that we are seeing more and more reverse image checkers and other sites that will do the quick work for you. Just simply reverse search images through Google, and they can point out information that tells you that the image is AI. The sites dedicated to just telling you if an image is AI or not can be hit or miss and there isn’t one that has stood out as being better than the rest, but they are still a useful tool when you need a quick answer or a starting point for telling if an image is AI or not. 

Misinformation and AI

The culmination of these AI images, and the issues with them, and the important ability to spot them is that we are seeing an uptick in AI images being used in misinformation and disinformation campaigns. 

Some of this is little simple misinformation like Katy Perry pretending to be at the Met Gala. But it also could be very serious stuff like the impact that deepfakes had on the 20204 election.

From accounts that post fake baseball news to news and story articles and sites that are fully and completely AI generated, the issues regarding the use of AI imagery and how they create disinformation and misinformation are many and in multiple forms. It emphasizes the importance of being able to tell what is and isn’t AI and how important that will be in determining what is and isn’t real information. 

Why is this important?

We focused a lot on how AI imagery can be used to create misinformation and disinformation in the last section, and I want to emphasize that this is probably the biggest and most important reason to be aware of and to be knowledgeable about  AI video and detection. I also want to focus lastly on just how prevalent and relevant AI imagery is, and why knowledge is important regarding this issue.

Most relevant to prevalence and misinformation/disinformation is the AI slop that our President has used in his second term. The fear that there are people who see and think this is real because it’s an official post by our Government is a real one. This isn’t the only example of this AI usage by governments.

Beyond that, knowing and not knowing what you are seeing if it is AI or not, is becoming a crucial skillset going forward. We are seeing AI images from sports teams, Hollywood films, billboards, and fast food advertisements. Just in the day of day of restaurants posting their menus, we are seeing AI, just look at these examples below.

Being knowledgeable and discerning about what you are seeing as AI or not is a problem that you will be facing more and more of, we are even now seeing AI actors and AI music. They are actively using disinformation to convince you that AI music is not AI music. 

This is an important skillset to be able to have in your toolbox. I know it’s easy to feel overwhelmed and overstimulated, but if you have made it this far, you have already consumed and are more knowledgeable about AI imagery than the average person.

Take what you have learned here and build a foundation that you can grow on and continue to add information and tools to help the fight against what is and isn’t AI.

This task is hard, but not impossible. With the right amount of information and tools discussed in this paper, you too will be able to tell if images are AI or not.