Elon Musk in a deepfake video.
It’s getting easier than ever to create deepfake videos.
That’s because the software to make these fake videos (which fool you into thinking President Obama is calling President Trump names or that Elon Musk is a baby) is widely available. Anyone can download the apps on the Dark Web.
And after someone superimposes the face of a president or celebrity onto someone else’s body (often in a pornographic movie), they can upload them to Facebook, YouTube or any other social media platform.
Once removed, they tend to reappear in yet another video somewhere else.
There is some gatekeeping using artificial intelligence, although other users outside of the platform are often the ones reporting the videos. But often times, the uploaders don’t tag the videos as deepfakes, leaving no way to know how many are online and which ones are even that troubling.
Sadly, the proliferation itself is hard to track. For a celebrity or politician, or even a friend from work, the videos are embarrassing because they look so realistic. No one is able to determine how many videos are out there or if the problem is getting worse.
Now a company called Deeptrace has found the answer.
Analyzing videos from around the Web, the company has determined there are almost 15,000 deepfakes in existence (14,678 to be exact), which is a significant increase from just 7,964 they found a year ago. That is a whopping 84% increase.
And the problem is getting worse.
One of the questions with gatekeeping on social media is that the companies housing the content are also proponents of free speech, so they don’t want to be perceived as blocking anything (even content that is unlawful and should be blocked).
There is a balancing act between allowing free expression and allowing the harmful material. At a recent tech conference, investor and book author Ellen Pao noted how the public square of social media has descended into a cesspool.
As she explained to the audience:
“I don’t think that’s an ethical decision. I don’t think that’s a values-driven decision. I don’t think that’s creating a good public square. I don’t think that’s doing a service for your users who are from the groups that are being hated on. I think you really have to think about your whole community. You have to think about the types of conversations you want to have.”
For now, deepfakes challenge the notion of speech. Many deepfakes are obviously harmful to the person depicted in the videos. They have no intent other than to cause embarrassment. And, the algorithms used by some of the social media companies can’t detect them.
Deepfakes will continue to proliferate, but the good news is that AI will eventually keep up, as Mark Zuckerberg noted last year.
Bots will roam around on social media platforms and will soon be able to detect illicit content like deepfake videos before they ever reach the masses.
Maybe the question we should ask ourselves then is who will control the bots.
