When I look at this photo of a seal and a puppy that somehow look like long-lost twins, my human brain gets the joke instantly. I can tell they’re two different animals, and I also get why they’re being compared. There’s context, humor, and visual nuance involved. But for a computer, that kind of recognition isn’t simple. Computer vision doesn’t work like human vision. We interpret meaning and emotion, while a computer just sees pixels, shapes, and patterns. Golan Levin’s essay really drove that home. Computers are not seeing the world, they’re processing data through whatever narrow lens we’ve given them.
To help computers understand what we want them to track, we use things like face detection, color tracking, optical flow, and trained models. These tools help narrow the field and make the computer’s “guess” more accurate. But still, it’s guessing. A puppy that looks like a seal might completely throw it off if the system wasn’t trained on edge cases like this. That’s part of what makes working with computer vision both fascinating and fragile.
In interactive art, computer vision opens up exciting possibilities. We can create responsive environments, playful installations, and performances that react to motion and presence. But the same tools are also used in surveillance and monitoring. There’s a tension between creativity and control that we can’t ignore. As an artist, I think it’s important to design with awareness. Just because the system can track someone doesn’t mean it should. I want to create interactions that feel intentional and thoughtful, not invasive. At the end of the day, I want the system to respond with care, not just accuracy.