Have you seen Paul Walker resurrected for Fast & Furious 7 or Mark Zuckerberg talking about getting "complete ownership of the stolen information of people around the world" or watched the moving apology of Jon Snow for the end of Game of Thrones?
If yes, then you've seen a deepfake. Deepfakes use a type of artificial intelligence called deep learning used to construct fabricated events - it's the 21st century's response to photoshopping!
Would you like to put new words in the mouth of a politician, appear in your favorite movie or perform like a pro? You can do that all with a deepfake. So what is a deepfake, exactly, and why are people so worried about it?
What Is a Deepfake?
Deepfake technology can insert anyone in the world effortlessly into a video or picture that they have never personally taken part in.
These technologies have existed for years, but it used to take entire studios full of experts more than a year to produce deepfake effects.
Presently, deepfake technologies, new automatic computer-graphics, and machine-learning systems can produce photos and videos much faster.
The term deepfake has become a blanket term to categorize everything from state-of-the-art AI-generated videos to any image that seems presumably deceptive.
In fact, sometimes what is considered a deepfake actually isn't. There is the notorious U.S. Democratic primary debate “crickets” video of Michael Bloomberg, which people were calling fake.
But it was actually just created with standard video editing skills, and no deepfakes were used in the production.
How Are They Created?
A creator must first train a neural network on real video footage of the person for several hours to create a deep-fake video of someone.
That will give them an accurate "understanding" of what he or she looks like from multiple perspectives and under different lighting.
And, to superimpose a duplicate of the person on a different actor, they will combine the trained network with computer-graphics techniques.
Although the inclusion of AI makes the process quicker than ever before, it still takes some time for this process to create a convincing composite that puts a person in a fully fictional scenario.
To prevent telltale blips and artifacts in the image, the maker must also manually adjust several of the trained program's specifications. The process is hardly simple.
What Are Deepfakes Used For?
Almost all are pornographic. In September 2019, the AI company Deeptrace discovered 15,000 deepfake videos online. Some 96% of those mapped faces were used in pornographic content.
And 99% of those images were of famous women projected onto porn actresses. As new techniques enable unskilled individuals to make deepfakes, revenge porn further in the entertainment industry might grow.
The Dangers of Deepfakes
Women will not be the primary victims of harassment. Deepfakes may well facilitate more common bullying, whether it be in schools or offices, as anyone can be put into fake situations that are absurd, risky, or threatening.
Companies worry about the role deepfakes might play in supercharging scams. In CEO scams, unconfirmed accounts of deep-fake audio have been used to swindle workers into transferring cash to scammers.
Extortion may become a case of significant usage. The greater worry about governments is that deepfakes pose a threat to democracy.
You can do the same thing to a senator running for reelection that you can do to an actress with revenge porn.
Bottom Line
The uncertainty surrounding deepfakes makes the technology even riskier. There could be many victims, but technology also provides protection so that people can do whatever they want and then call it fake.
There's a lot of hope in new technology, but sometimes all this advancement is downright scary.
