“Generative AI Has a Visual Plagiarism Problem”
As a consequence of using benchmarks intended to measure human performance generative models are designed to memorise, which makes them prone to overfitting the output. There is a wealth of academic research demonstrating this