The War on What's Real
“The more techniques you develop to distinguish fact from fiction, the better the fiction becomes,” explains Jon Brandt, Director of the Media Intelligence Lab at Adobe Research, and the first AI researcher the company hired 15 years ago.
At Adobe, Brandt’s job isn’t to create future products but to spearhead and coordinate the bleeding edge IP with which they’ll be infused. For more than a decade, Adobe has owed many of its improvements to AI. In 2004, it introduced its first AI feature with automatic red-eye removal. Then face tagging in 2005. Now AI powers dozens of features from cropping and image searching to lip-syncing.
For the most part, these are handy updates to Adobe’s creative tools, gradually updating the features of its apps. But knowing where my questioning was leading him–that Adobe is the biggest, most profitable company in image manipulation, and single-handedly ushered in the reality distortion field of Photoshop that we all live in today–Brandt doesn’t mince words.
“The elephant in the room is Sensei,” he says. “It’s Adobe’s name around how we are approaching AI and machine learning for our business. It is not a product–which is frequently a source of confusion–but a set of technologies.”
Launched in 2016, in a coincidental parallel with the rise of fake news, Adobe’s Sensei projects makes Photoshop’s clone stamp tool look like a janky old VHS tape. Thanks to machine learning–and a small army of research interns Adobe recruits every summer to publish work at the company–Adobe can create a doppelgänger of your own voice, to make you say things you never said. It can automatically stitch together imaginary landscapes. It can make one photo look stylistically identical to another photo. And it can easily remove huge objects from videos as easily as you might drag and drop a file on your desktop.
These research experiments–which to be clear, haven’t been rolled out into Adobe products–aren’t just powerful media manipulators. They’re designed to be accessible, too. They don’t require years of expertise, honing your craft with artisanal tools. With the infusion of AI assisting the user, the apprentice becomes an instant master. Soon, Adobe plans to make image manipulation as simple as talking to your phone.