An essential feature of Generative AI that does not attract public attention (and publications) is that, unlike systems designed for recognition, function approximation, etc., there are no objective “correct answers” to compare system output. Accordingly, there is no way to calculate the percentage of correct answers or the error level;