I like to think of myself as internet savvy with high media literacy. Still, last week while scrolling Instagram I spotted the most adorable photo of a baby seal eating something out of a person’s hand. I flipped my phone around to my teammate Sam and showed her.
She immediately responded, “You know this is fake, right?”
Gulp.When I took a closer look, the signs were pretty obvious.
Listen, we’ve been the subject of poorly written AI articles so I feel like I’m already on high alert but I was fooled by the cuddly creature on the screen. My brain wanted it to be real so I missed so many red flags. While a silly fake photo of a seal seems harmless, I think it’s important to stay vigilant about the ever-changing world of AI.
Here are some ways to make sure you aren’t being fooled (like me):
Sleuthing Time
AI is a powerful tool that can be used for good. We’re hearing of so many medical advances developing using AI technology. But right now I want to focus on what we’re seeing online: clickbait articles, fabricated images, and “deep fake” videos. We will share tips on how to spot AI through these three mediums.
That Image Seems Off
Comic artist Peter Chiykowski recently tweeted a few tips for spotting fake images. He encourages people to right-click and open the image to zoom in and check closer for these things:
Consistency: Is there a symmetrical pattern in the image? See if it is actually continuous.
Spelling: Are there words written out? AI struggles with writing text.
Fingers: This is a funny one but when hands show up in AI images, they tend to have 6+ fingers.
Was This Article Written By A Robot
We’ve been the subject of some obvious AI-written articles. We read them aloud and are crying with laughter at how BAD they were. There was one that started every sentence with “Let me tell you…” It was hilariously awful.
The people over at Originality shared some helpful hints for recognizing when a blog post may have been written by a computer.
Patterns or repetition: AI tends to repeat sentence structure.
Transitions: AI-written articles seem to abruptly switch between topics and ideas.
Facts: The most popular available AI program is trained on data before 2021. Anything after that is going to be factually incorrect.
Unnatural movements: Not shockingly, computers are bad at mimicking human body behavior.
Inconsistencies: The light, shadows, and video quality changing within a video is a good sign it’s been tampered with.
Lips don’t match audio: If the video includes talking, this may be a big red flag that you are watching a deep fake.
Do Your Research
One model of media literacy, created by research scientist Mike Caufield, is called SIFT. That stands for four steps: Stop. Investigate the source. Find better coverage. Trace the original context. The overall idea is to slow down and consider what you’re looking at — especially pictures, posts, or claims that trigger your emotions.
If you are reading a blog that may seem off, take a look into the author’s background and see if they are a credible source.
Is it a photo you can quickly download? Do a reverse image search or run it through an AI detection tool, like one of these from a Forbes top 10 list.
Read the comments. Is the article/photo/video posted on social media? Most likely someone in the comments has shared their doubts about the validity of the post.
Last but not least: If you are not sure if it’s fake, don’t like or share it.
This NPR article shares a few fake photos that went extremely viral, one of the Pope wearing a puffy jacket. It looks extremely realistic and most people could only tell when they zoomed in and saw that he wasn’t actually holding the coffee cup in his hand.
Of course, while the Pope wearing a puffy jacket seems harmless, the increase in edited images of public figures is extremely problematic. Don’t be part of the problem by sharing an image or post that has no basis in reality. Increasing your media literacy is vital to being a citizen on the Internet. And if you see a video of me saying I’m an extrovert who loves to go to parties…it’s fake.
