It is

Neural Networks and by extension LLMs are simply statistical models reorganizing training data based on statistical probability. It’s not creating anything new, it’s taking the inputs and more or less “translating” it.