It's no more insightful than using Markov Chains. Clever algorithms blend up pre-existing combinations of words and spits them out. Once in a while those words

It's no more insightful than using Markov Chains. Clever algorithms blend up pre-existing combinations of words and spits them out. Once in a while those words will make sense and might even apply to a useful idea previously unthought of. GPT-x increases the likelihood its output will at least make sense to a human reader. It still requires a human to determine if its output is useful. The program itself has no practical way of determining if its output (an "idea") is meaningful or useful, let alone novel.