We’ve lived in the age of Photoshop for a long time–even North Korea wields the tool like a weapon of propaganda. But Adobe keeps building tools that up the ante on what’s possible, leading us toward a world where nothing we see seems certain at all.
Over the past two years, Adobe has ventured deeper into artificial intelligence with initiatives like Adobe Sensei, the AI the company is building to power many of its tools. In 2016 it demoed an experiment that let you literally put words into someone else’s mouth (some unrelated researchers took the idea even further and faked a speech by President Obama). And this week at Adobe’s yearly MAX conference we saw even more AI experiments (the company calls them “sneaks”).
Two of the demos are downright unsettling–they let you add or remove virtually anything from a landscape, filling the gaps automatically. Even in video.
The first project is SceneStitch, and it’s built to replace large sections of photos. It’s an advanced riff on the company’s existing Photoshop tool, Content Aware Fill, which lets you highlight a portion of a photo and ask the computer to essentially clone stamp it smooth, generically filling the hole with the scene around it. Using AI, SceneStich goes a lot further. It lets you highlight a part of a photo–say, a road in a desert–and replace it with all sorts of plausible objects from real photos; maybe a boulder, or a pond, could go there instead. The brilliance is that, while SceneStitch’s AI suggestions aren’t even necessarily good, the human in the loop can easily pull the decent ones out–all with zero talent or effort. In the video above, you can watch as SceneStitch deletes a whole neighborhood from the cityscape of Denver, replacing it with a harbor full of boats. It’s like a one-button gentrification filter.
The second project is called Cloak, and it’s aimed at removing things from full motion video. Adobe put it to work on what looks like a traffic camera that’s erected in front of The Cathedral of Christ the Savior in Moscow. The tall pole destroys the view of the picturesque architecture. But just by tracing a mask around it, the user can set Cloak on making it disappear, filling in every frame with the missing bits of the cathedral. It’s downright spooky to watch, like a god erasing the existence of an object with a snap.
Since both projects are “sneaks,” they’re only prototypes that aren’t planned for any specific product yet. Even still, they continue to demonstrate a principle that’s really quite scary in the digital age. Photoshop, powerful though it is, required real time and skill to master. Likewise, the video altering capabilities of After Effects are known to a relative few experts.
But Adobe is getting so good at building image manipulation tools that it’s effectively removing human skill from the equation, making fakery available to anyone through ingenious software and user-friendly UI. And while that’s textbook-level good design, in the era of fake news, it might not be as good for society at large.