In the last column, we started discussing large language models (LLMs)-artificial intelligence programs like ChatGPT, Google's Bard, and Bing Chat that can generate human-like answers in response to prompts. I mentioned that lawyers should be careful when using them-especially given their disturbing habit of making up authorities to support their answers! Before the ink was even dry on that column, one New York lawyer found out about LLMs' limitations the hard way. He asked ChatGPT to help him find case law that would help his client pursue a federal claim against an airline. ChatGPT obliged, providing him with eight seemingly relevant airline cases, which he cited in a brief to the court. But there was a problem: the cases weren't real. The A.I. had invented them. Unfortunately, the attorney never verified the cases, leading to an angry judge and possible sanctions. So if LLMs aren't quite ready for comprehensive research and analysis, where can they be helpful for a legal writer? Here are a couple of ideas.