“We are not in AI winter anymore. It’s late spring,” declared Dazza Greenwood, researcher at MIT’s media lab and executive director of law.MIT.edu. In real time at BYU Law’s Future of Law forum, Greenwood used a student inquiry about using AI to amend a patent to demonstrate how detailed inputs can generate astonishingly accurate results. The AI model proposed the very changes that had been made to avoid a patent lawsuit.
Historically, law has not been an ideal space for AI because results were generated in a logical, decision-tree flow, while legal answers are nuanced and require judgment. “Now what we have is well-suited to the legal domain,” Greenwood said. The key is to give the AI model as much context as possible to yield the most accurate results. Greenwood recommended “utility prompts” that give specific orders and fine-tune the results by specifying a purpose and the desired formatting. “I take great joy in creating enormous prompts!” He urged students to use AI to validate understanding and brainstorm ideas, but he warned, “Do not copy and paste results—ever.” Assessing AI-generated results for accuracy is essential and requires expertise. This necessity, Greenwood observed, is good news for law students: “Law school is still really important because the legal world needs capable shepherds.”
Comments (0)