The article is primarily about one particular machine learning powered device that is making a lot of mistakes. It explicitly includes the above paragraphs to explain that it’s not talking about LLMs for the most part (though it does start to mention their increasing use later)