Oddbean new post about | logout
 Researchers have developed an AI program that can predict the chemical composition of salt stains by analyzing photos. The program was trained using 7,500 photos of 42 different types of salt stains, which were translated into 16 parameters capturing features such as deposit area, compactness, and texture. The machine learning algorithm, once trained, was able to correctly identify the composition of salt stains in 90% of attempts, even for images that were not part of the initial dataset