Cameras are related to being able to see. It does make sense. But maker Diego Trujillo Pisanty came up with a different idea. He used a Raspberry Pi to make a camera that can't see. To be clear, it's not a camera to help the blind, but one that can't see, at least when it comes to its visual senses. Instead of a lens, it uses sound to figure out what is going on around it. The camera is made with a Raspberry Pi 3B, but it can also be made with a Pi 4. The project is meant to look as much as possible like a point-and-shoot camera, but it will make pictures from sounds. Point the horn, press the button, wait while the picture is being made, and then look at it. Inside the body is a small LCD.
It works by training AI, which is why the pictures it takes are in color. This isn't just an echolocation system. It also makes a very clear point about how AI is trained and why it can be helpful for an artist to build their own AI that focuses on artistic purpose. Diego says, "The code is art just as much as the object and the images that come from it." As Diego Trujillo Pisanty goes on to explain, the Artificial Neural Network (ANN) has a lesson that could serve as a warning for all of us. His ANN was trained on the sights and sounds of Mexico place, so it favors the place it 'knows.' If you take the 'Camera' out of the city, it will make pictures that look like Mexico City. "Everything is a city to the machine," he says. There is also a clear worry that places with more investment in AI might have a "sterilizing" effect on the rest of the world's society.
Jean Dubreil