I realise this but if the developers can't identify English mistakes in the interface how can they identify how good the AI's advice is? I also have examples of AI giving bad advice about language and I plan to make a list of those too.