If you’ve been listening to the news or engaging in conversations about artificial intelligence (AI), you’ve likely heard the term deepfake. But what exactly is a deepfake?
Definitions
Merriam-Webster defines a deepfake as “an image or recording that has been convincingly altered and manipulated to misrepresent someone as doing or saying something that was not actually done or said.” In the Oxford English Dictionary, it’s referenced as “any of various media, esp. a video, that has been digitally manipulated to replace one person’s likeness convincingly with that of another, often used maliciously to show someone doing something that he or she did not do.” According to the Government Accountability Office, it’s “a video, photo, or audio recording that seems real but has been manipulated with AI.” In short, a deepfake is some sort of media (a video, image, or audio recording) that sounds or looks real but is actually fake.
The Importance of Being Informed
In a world where people believe what they see or hear, this type of transformative technology is poised to have a significant impact on society. It’s already being used to produce fake news stories, to create clickbait content, and to mislead voters. In a democracy where citizens must be able to make informed voting decisions, it is imperative that we all become digitally literate in this new technology.
Four Common Types of Deepfakes
The first step in developing deepfake literacy is to understand the basics, including what types of deepfakes are most commonly being produced. The following are the four most popular deepfakes techniques.
This is exactly what it sounds like: One person’s face is swapped out or replaced by another face in a video or picture. Some of these are created by seasoned pros, and others are made by average consumers who have downloaded apps with this capability. People use these apps to put their faces on images of their friends or celebrities. Most people are creating a face swap to have fun.
The YouTube Channel Ctrl Shift Face has a collection of video face swaps meant to entertain. There’s a very convincing one that has Jack Nicholson’s face swapped out for Jim Carey in a famous scene from The Shining. If viewers didn’t know that Jim Carey wasn’t actually in that film, there’s a high likelihood they would believe that this footage is real. It looks that convincing. A deepfake of a movie clip like this probably won’t cause any major harm, but consider if a face were to be swapped out in a photo or video from a crime scene or at a political protest. In those contexts, the misleading image could easily damage a reputation or be used to mislead an audience. This type of misinformation is a concern as the United States heads into election season.
In this case, one person’s face is not replaced by another’s; instead, a video of someone is manipulated to make it look like they are saying something that they never actually said. For example, there is a famous YouTube video, This is not Morgan Freeman – A Deepfake Singularity, which shows an AI deepfake of the actor demonstrating how real it can appear. It starts with the face of the actor on the screen, saying, “I am not Morgan Freeman, and what you see is not real.” While the lips don’t exactly match the words being voiced, it really does sound like Morgan Freeman. Someone who was not paying close attention could easily be convinced that it was, indeed, Morgan Freeman who was speaking.
Similar to the worries about face-swaps, there has been increasing concern that face manipulation might also be used to spread election misinformation. Candidates for office might be made to appear that they are saying something that they never really said. Because a video is showing them saying it, it can be very convincing, and viewers may believe that it’s real.
This type of deepfake can take a couple different forms. On the one hand, it might mean simply converting text to speech, like with the Google Chrome extension Speechify. This can be a really great accessibility tool for people who struggle to read. It can also streamline commercial production projects. Companies often use this type of technology to produce AI-generated voice-overs for training videos.
Another use of voice synthesis is to reproduce a person’s voice. That reproduction can then be used to make it sound like the speaker is saying virtually any message that can be typed into the processing program. Hollywood recently made use of this type of voice synthesis in the movie Top Gun: Maverick. The actor, Val Kilmer had lost his voice due to throat cancer, and this technology was used to allow him to speak in the movie. While he wasn’t really speaking, the synthetic version of his voice made it sound like he was—and it was convincing.
While these are positive, legitimate uses for this type of AI, it can be used for nefarious purposes as well. In fact, this happened during the 2024 New Hampshire primary elections when a fake robocall went out using President Biden’s voice to discourage voters from going to the polls. It sounded very authentic and was used to impact the outcome of the election.
Within the last year, dozens of AI image generators have been released, and many of them are free. These include tools like Adobe Firefly and OpenAI’s DALL·E 3. Users type in a description as a text prompt, and the tool produces several versions of an AI-generated image. These images are created based on patterns identified in millions of images that have been scanned and used to train the AI generator in how to construct a new image based on those patterns.
This image generation process has caused concern on multiple levels. Many photographers and artists who have posted their images online believe their work has been used to train these AI models without their consent. They believe that these AI companies are essentially stealing their unique styles and techniques to create these new images. To offset this concern, some companies like Adobe are only using images for which they own all the rights.
Recently, deepfake images have been used to pass off a fictional scene as something that really happened. Compromising deepfake images of Taylor Swift went viral on the social media platform X, being viewed 47 million times before X took them down. In other examples, there was a deepfake image of former President Donald Trump getting tackled and arrested by New York City police officers and one appearing to show Vladimir Putin in prison. The images were fake, but they looked real enough that they got circulated on social media with little or no context, leading some people to believe that they were real.
The proliferation of these four types of deepfakes is growing quickly. A 2023 Reuters article about deepfakes cites that from May 2022 to May 2023, the publication of video deepfakes tripled. The report also states that there were eight times as many voice deepfakes in 2023 as there were from the previous year.
While this technology is impressive and powerful, it is also potentially problematic. Scholars are wondering what will happen in a democracy where voters aren’t sure what they see or hear. How will they make informed choices at the ballot box?
Those questions are still to be answered, but most people agree that this technology is not going away. In fact, it’ll likely continue to get more sophisticated and harder to detect.
Therefore, it is critical that we all become savvy digital consumers, and the first step in doing that is to know that this technology exists, to have a basic awareness of how it works, and to begin viewing and listening with a discerning mind.
AVID Connections
This resource connects with the following components of the AVID College and Career Readiness Framework:
- Instruction
- Rigorous Academic Preparedness
- Student Agency