Introducing Paragraphica, a sensorless camera, powered by AI

A Danish designer has developed a camera that, thanks to artificial intelligence, is able to take pictures without a lens, based only on geolocation information.

Paragraphica is modeled on a type of star-nosed mole. I’camera It exists in a physical form but can also be used online, without a certain number of parameters that can be modified with the object.

Interpretation of the place

Paragraphica is still in prototype status, and has already had a modicum of success online. The result of the work of Danish designer Bjørn Karmann, it is an original-looking camera that draws its attention from the use of artificial intelligence to deliver original images. It does not crack the technology codes, as it uses a Raspberry Pi 4 and is based on the Noodl program, the Python language and the Stable Diffusion API. The odd shape placed in place of a real camera sensor was inspired by the nose of Star Condylure, a mole that has evolved its nose to smell and feel its surroundings, without having to use its eyes.

This is the general idea behind this concept camera, according to its creator. “As AI language models become more self-aware, we will also have a limited imagination of how they see the world. The camera provides a way to experience the world around us, a world that is not limited to visual perception alone.”
In short, the image sent by the device will be an interpretation based mainly on the available data.

A working prototype

So the model developed by the designer will use the location data of the person using the camera. The device will collect geolocation data using open APIs (a big question in the explosion of AI, one Reddit decided to answer by accessing its own API push), be it address, weather, time, or nearby places. The system then outputs a short paragraph summarizing the situation on screen, before converting it into a generated image.

Paragraphica first displays a paragraph (hence its name) that summarizes the data obtained and then converts it into an image.© Bjorn Karmann

The resulting “image” is not just a snapshot, but a complex and accurate reflection of where we are, and perhaps how the AI model “sees” that place. Bjorn Karmann tells us on his website. The man explains that the AI seems to be able to pick up on the feelings or moods that arise from a place, but in a strange way, because the images don’t look exactly like the real place. It is also possible to try a prototype directly onlineby allowing its location to be shared.

The guy is still designing an actual device with physical buttons to adjust several parameters that will affect the final image.