We know, in a world of uncertainties, AI is coming for all of us in different ways. Trying to keep up with all the changes (and I am not even mentioning our overseas adventures) is exhausting, overwhelming, and frustrating. How to cope? More reliance on AI?
The Wall Street Journal recently ran an article comparing the three large learning machines (Claude, Gemini, and OpenAI) in a kind of LLM legal writing Olympics.
The results were fascinating. Each of the three competitors was better in some ways, and worse in others. Each bot had quirks of its own. How to tell a bot from a human?
In this admittedly unscientific test, one way to tell a bot from a human was vocabulary. If it sounds like it’s “a panicked college freshman trying to sound profound,” it’s a bot. If the article, memo, or document starts out by telling the reader what it’s about, it’s a bot.
All three bots hedged, reluctant to give opinions. “On the one thing … on the other.” That wishy-washy language is not what clients are paying for. They are paying for our opinions and our advice with available options about how to proceed. Clients want clear directions and advice; save the erudite for law review articles.
The time will come, sooner rather than later, when bot writing will be essentially indistinguishable from what we humans write. It’s about to become a lot more difficult to choose the real from the artificial.
You are not a bot, so don’t write like one. Clients do not want to read (or pay for) pages and pages of legal gobbledygook that, in the end, only confuse the reader while the meter runs. Perhaps for law review articles and other scholarly compositions, more is more, but for the everyday lawyer who is just trying to KISS (Keep It Simple Stupid), twisting yourself into a legal literary pretzel does no one any good, especially the reader. Get to the point quickly before eyes glaze over and the reader snores.
On another AI topic, is a lawsuit really final even if it’s been settled and the case dismissed with prejudice? No, not according to ChatGPT, a font of legal (mis)information (ahem).
Nippon Life Insurance has sued OpenAI in federal court in Chicago, alleging that OpenAI engaged in UPL, that is, the unauthorized practice of law. The basis? ChatGPT advised the settling plaintiff in the underlying disability case that she could reopen that dismissed lawsuit. (She had a case of settler’s remorse, not that any settling party has ever felt that way.) Nippon’s complaint alleges that ChatGPT is not an attorney and therefore cannot give legal advice.
The plaintiff thought that her attorney (a human, not a bot) had given her bad advice about whether she could indeed reopen the dismissed case. So, she went “attorney shopping” and looked to ChatGPT for advice. Guess what? ChatGPT told the women that indeed she had been given wrong advice. The woman fired her counsel and looked solely to AI for advice and moved to reopen the closed case. After that was denied, she filed a new case and dozens of motions allegedly using AI again, including a hallucinated case. OpenAI says that Nippon’s case lacks merit. Really? Who is responsible for a bot’s conduct? Certainly not the bot, at least not so far.
On how many levels is this scary? Let me count some of the ways. UPL is a big problem for bar disciplinary agencies. Too many nonbarred peeps in the field. How to enforce UPL against a bot? That’s trying to nail Jell-O to a tree. How could the disciplinary process be used to outlaw the use of AI? Should it? How can lawyers protect themselves, if at all, from AI dissing their advice resulting in an unhappy client who fires the lawyer and then files a complaint with the bar based on that allegedly bad advice? Which, in this case, was correct advice? How does the court order a bot to pay a Rule 11 sanction? Is your head spinning yet?
Reliance on incorrect information from ChatGPT or any other bot that leads to frivolous lawsuits, both in court and in unjustified bar discipline cases, only makes the legal system grind ever more slowly and lead to even more crap filings. Is reliance on a bot merely general legal information or specific legal advice?
Pass the Pepto, please. Or an Excedrin. Or maybe both. Perhaps a bot can suggest what to take.
Or would that be practicing medicine without a license?
Jill Switzer has been an active member of the State Bar of California for over 40 years. She remembers practicing law in a kinder, gentler time. She’s had a diverse legal career, including stints as a deputy district attorney, a solo practice, and several senior in-house gigs. She now mediates full-time, which gives her the opportunity to see dinosaurs, millennials, and those in-between interact — it’s not always civil. You can reach her by email at oldladylawyer@gmail.com.
The post Bot’s Not Nice appeared first on Above the Law.