Called Photoguard new device developed at the Massachusetts Institute of Technology Boston to make life more difficult for artificial intelligence, which is increasingly being used to edit photos. This new technology creates a kind of pixel-level disturbance, something invisible to the human eye but capable of confusing the algorithms that have to recognize what each part of the image represents. As a result, it will become almost impossible to create manipulations that often radically change the meaning of a photograph.
Artificial intelligence can greatly facilitate and speed up the work of editing any image. Take Dall-E or Stable Diffusion for example, which start with an image uploaded by the user and then take action. following text instructions often with sensational results, just like the tools Adobe has introduced in recent months for Photoshop. However, there is a way to put a digital lock on a photo and prevent the AI from manipulating him and it’s called PhotoGuard. The tool was developed at MIT CSAIL i.e. Laboratory of Computer Science and Artificial Intelligence belonging Massachusetts Institute of Technology and operates at the invisible level of a single pixel, alternating key pixels that are analyzed by AI to “see” what the image represents.
These digital artifacts are spokes in the wheels of algorithms that cannot act correctly, which affects their effectiveness: the modification will be taken out of context, of low quality and not very reliable, revealing at first glance. “Although I’m happy to help find a solution to this problem – commented the doctoral student Hadi Salman who led the project – on the other hand, I realize that much work needs to be done to make this defense practical. Companies developing these models must invest in developing robust immunizations against possible threats posed by these AI tools.“.