I can imagine the response including questions. Eg “do you mean X or Y, here are the implications”. That way the machine gets better at helping us clarify our thinking. Right now it’s just blurting out a response that fits your instruction with no attempt to clarify.