tailcalled comments on ChatGPT’s “fuzzy alignment” isn’t evidence of AGI alignment: the banana test