Subscriber Benefit
As a subscriber you can listen to articles at work, in the car, or while you work out. Subscribe NowLawyers from around the state learned strategies and techniques for making the most out of artificial intelligence at the Indiana State Bar Association’s annual summit.
Sara Kubik, owner and founder of Kubik Legal, a Crown Point-based legal firm, shared tips and tricks for firms using ChatGPT, a generative AI software developed by OpenAI, at one of the state bar’s Friday breakout sessions.
From how to write a successful prompt to brainstorming marketing materials, Kubik pointed out the ways that the sophisticated software could make daily tasks flow more efficiently.
“I’ll ask ChatGPT to rephrase or summarize something,” Kubik said. “It could be something as simple as an email, especially to a client, to change the tone of your message.”
Alongside being a user of AI in her legal career, Kubik, who has a P.h.D in technology, is also an “AI trainer,” working with large technology companies to educate them on how to improve their AI models’ responses to legal questions.
“I am literally acting as a teacher,” Kubik said.
Kubik cited an example where she explained, she will create a prompt, focused on a specific legal question, and then evaluate the AI’s answer. If she sees inaccuracy in the response, she will document what it should have been and provide the company with steps to get the correct answer.
This practice is known as Reinforcement Learning from Human Feedback—essentially, it’s “bringing humans in the loop,” she said.
Although it is taking in human expertise, Kubik emphasized to the bar members that AI models like ChatGPT have “no legal reasoning.”
“We talk about it, ‘Oh, it’s reasoning, it’s thinking’; It’s not. It’s predicting the next word,” she said.
Although Kubik said she thinks the software is amazing, she also said she doesn’t think it would replace lawyers.
“When it comes to technology, this stuff is, it’s enormously powerful,” Kubik said. “They can enormously screw up.”
In 2023, a federal district court judge fined two attorneys after they submitted fake case citations in their court filings in Mata v. Avianca, which had been generated through ChatGPT.
“Technological advances are commonplace and there is nothing inherently improper about using a reliable artificial intelligence tool for assistance,” wrote New York Southern District Judge Kevin Castel in an opinion in the case. “But existing rules impose a gatekeeping role on attorneys to ensure the accuracy of their filings.”
Because of the danger these “hallucination” cases could bring, Kubik emphasized to the crowd the need to “verify everything.”
“We’re lawyers, we have our law license to consider; ChatGPT doesn’t care about that,” she said. “We also have ethical issues that we need to follow; ChatGPT doesn’t care about that, so we’re always going to verify.”
So, Kubik encouraged users of the software to go “back and forth” with the model, redirecting it when it responds to a prompt, and even telling it when it’s wrong.
Please enable JavaScript to view this content.