Deepfakes were going to be artificial intelligence’s first great scourge. According to journalists and experts in the last two years, the technology that automatically stitched and animated one person’s likeness onto another person’s body would be used by malicious actors to create torrents of fictitious speeches from world leaders and public figures.
But it seems like that was mostly wrong. Certainly, individuals have been hurt by deepfakes: In 2019, a report emerged that more than half of deepfake videos online were related to porn, targeting the women that the algorithms were originally designed to emulate and harass. But the tech has not (yet) shook the very meaning of truth online.
Instead, tech companies are starting to cash in on the technology that powers deepfakes. Last month, Snap bought a San Francisco-based startup for a reported $166 million. AI Factory had previously partnered with Snapchat to capture 3D models of user's faces and plaster them onto about 150 scenes in a feature called “Cameo.”
The images produced by Cameos were slightly cartoony, meaning that the digital avatar looked a little bit more like a Bitmoji than the user’s real face. But the effect was essentially the same — it took someone’s likeness and spliced it onto another body or into another scene.
And it seems other tech companies are considering integrating deepfake-style technology onto their platforms, too. A recent report suggested TikTok is creating deepfake-type software as a kitschy tool for users to put their faces on other bodies.
The tech has not (yet) shook the very meaning of truth online.
If it wanted to, Apple could roll out a similar product. The core technology behind Apple’s Face ID creates a depth map and infrared image of your face. The company says those images are accurate down to the millimeter, meaning the base image to recreate your face is likely already on your iPhone; this same technology powers the “Memoji” characters on iMessage and FaceTime.
Shiny new filters like Cameo can breathe new life into aging social media companies. Snap, beleaguered by falling monthly users and endless copycatting by Facebook, more than doubled its monthly app installs when its “gender-swap” filter went viral in May 2019. A new filter that puts a user’s face onto an iconic scene from a movie or someone else’s viral video could be a leg up on a competitor. Snap did not respond to a request for comment regarding how the Cameo tool would be used in the future, or how it differed from deepfakes.
But how social media companies implement the technology is drastically important. These product decisions, made with no government or user oversight, could be the difference between our worst fears about deepfakes actually playing out and a two-week bump in Snapchat’s daily active users. Any of the platforms listed above could release a deepfake feature that puts any scanned face on any body in a video tomorrow.
That doesn’t seem to be likely. The Cameo feature inside Snapchat played it safe by only allowing the cartoon faces to be placed on a select number of videos originally created by the company. TikTok’s feature also allegedly was built around pre-selected videos that ByteDance, TikTok’s parent company, has the rights to use.
This means that the immediate threat of deepfakes is changing. It’s no longer a matter of who will gain access to a sophisticated A.I. tool, but how platforms encourage or shape the use of the technology with product decisions.
It also means every platform will have to take a stand on the technology. Some, like Snap and TikTok, look like they will embrace it. Meanwhile, Facebook recently tightened its policies around manipulated media, even if the platform still allows parody and satire. For now, though, our worst nightmares about deepfakes have yet to come to fruition.