Not only that but LLMs unsurprisingly make similar distributional mistakes as humans do when asked to generate random numbers.