Скачать или смотреть A blind camera transforms sounds into images thanks to AI

➡️ Read the full article here: https://tinyurl.com/zxr4d6wr

An AI-powered "blind camera" that can convert sound into image? You won't believe it until you see it.

This camera's inability to detect light is a stroke of genius that demonstrates the shortcomings of AI training.

The ability to see is intrinsically linked to cameras. It is logically sound. However, designer Diego Trujillo Pisanty had another idea. The camera he built with a Raspberry Pi was blind.

This is not a camera designed to aid the visually impaired; rather, it is one that lacks any sort of visual perception whatsoever. It doesn't utilize a lens, but rather its hearing to interpret its environment.

The camera is built on a Raspberry Pi 3B, though a Pi 4 might be used instead.

The finished product is supposed to resemble a point-and-shoot camera as closely as possible, but it will create images based on audio. Aim the horn, depress the shutter release, and examine the resulting image. The body conceals a little LCD screen.

The color photos it takes are a result of the AI training it uses. You can't just call this an echolocation device. It also provides a crystal clear case for why an artist could find it useful to train their own AI specifically for creative tasks. Francisco "Diego" Diego once remarked, "The code is art just as much as the object and the images that come from it."

As Diego Trujillo Pisanty elaborates, we might take a warning from an important lesson learned by the Artificial Neural Network (ANN). His ANN was conditioned on data from Mexico, so it prefers to stay in the environment it is most familiar with. The 'Camera' will produce photos that look like they were taken in Mexico City, even if you take it outside of the city.

"To the machine, everything is a city," he explains. Investing heavily in artificial intelligence may have a "sterilizing" influence on the rest of the world's society, which is a legitimate concern.