From the cute visuals of the first full-length CGI animated movie (Toy Story in 1995) to special effects so realistic (I’m looking at you, Ex Machina, an Academy Award winner for Best Visual Effects Movies), it can sometimes be nearly impossible to knowwhat’s real and what isn’t. But that’s the unfortunate reality of modern technology: we create powerful tools that can be used for both good and bad. Welcome to the shady world of deepfakes.

What is a Deepfake?

Currently on Merriam-Webster’s Words We’re Watching list, deepfakes are described as: “a video that has been edited using an algorithm to replace the person in the original video with someone else (especially a public figure) in a way that makes the video look authentic.”

Gee, what possibly could go wrong?!

Sometimes, deepfakes are created purely out of fun, like the whole TikTok channel dedicated to Tom Cruise deepfakes. But sometimes, there is an agenda behind deepfakes to sow confusion, distrust, and doubt. For example, there are deepfake videos on both sides of the Ukrainian conflict that fall in that category. Because of real-life threats like this, schools like the University of Washington have developed entire courses dedicated to identifying all sorts of misinformation in online sources. It’s definitely not as easy as it used to be. 

Is It a Deepfake?

Norton, maker of a very popular antivirus software, has a fifteen point list around how to identify deepfakes. Unnatural body motion and weird postures are on the list, but so are things like:

• Teeth that don’t look real. (Are individual teeth visible?)
• Unnatural eye movement. (Did they blink?)
• Misalignment of audio and video. (Does it look like a poorly dubbed old spaghetti western movie?)

Interestingly enough, the technology behind NFTs is something that could help clear up whether an image or video is a deep fake by keeping an immutable record of the origin of the files in question. When the image and all its metadata are stored in a blockchain, any further changes will be explicitly stored such that anyone can see the history of who, what, where, and when the image file was created and added to the blockchain.

Artificial Intelligence (AI) is another area that’s being explored to help control the deepfake problem. Several of the big tech giants are working on having AI identify deepfakes. Ironically, it’s AI and its sibling, Machine Learning, are what often generate the deepfakes in the first place. This adds a whole new dimension to cyberwarfare and computers fighting computers.

Pop Quiz!

So, think you can figure out which images are deepfakes and which aren’t? Microsoft has a fun online quiz you can take to see how well you do in identifying these types of images. Or, for even more fun, you and your friends can work through an escape room game called the Euphorigen Investigation, a project co-developed by the CIP, the Technology & Social Change Group at the UW Information School, and other partners.

Why We Care

The existence and growing prevalence of deepfakes is a huge problem. When it becomes too hard to figure out what’s real and what’s not, the safest choice is to assume it’s all not real and nothing can be trusted. People cannot make informed decisions when there are no trusted sources of information. Even where deepfakes are relatively obvious, their existence sows general distrust in the information available. 

There are a variety of fact-checking sites out there (here’s a good list) that can help, and it’s worth taking a moment both to find a reputable fact-checking service AND to make sure it has solid ratings with regards to political neutrality.

Good luck out there, and don’t forget to check your sources!