Artificial intelligence (AI) has come a long way since the emergence of deep learning, or deep learning. AI relies on neural networks to learn to perform tasks on its own, with only one goal to reach and no other information. With regard more specifically to images, this allows the creation deep fakesor to generate images from simple sentences as they do Dall-E from OpenAI, google image either halfway.
However, to achieve these results, the researchers must train the algorithms on the databases. In the case of images, the largest public database is called Laion-5B and contains 5.8 billion images with associated texts. In theory, all images come from public, royalty-free sources. In reality, this is not always the case.
Images often used without consent
A new site called Have I been trained? allows you to search this database. This tool was designed so that artists and photographers can tell if their works have been used, but the tool also lets you know if your own photos are included. Simply search by entering your name or using your photo. The site ensures that the images used for research are not saved.
Certain elements of the database are particularly problematic. A woman In particular, he found photos taken by his doctor after a series of jaw surgeries, photos that would never leave his personnel file. The good news is that it’s easy to find out where the photos are from. Laion-5B does not contain images directly, but links to them. A simple right click allows you to find out which site they are on. You will then need to contact the site administrator to request removal.
Read also in Future