The Results of Training an AI on 2.5 Million Fart Samples

A humorous post details an experiment where a neural network was trained on 2.5 million fart samples collected over three months. The creator claims the AI achieved 87% accuracy in identifying diets based on a single sound, categorized seven basic fart types, and even generated 'fart symphonies.' The experiment also yielded a 52% accuracy rate in stock market predictions. However, the most notable outcome occurred when the AI was tasked with creating 'the perfect fart,' resulting in a 3.7-second sound that allegedly caused the creator's dog to cry and their roommate to move out.