AI in the Legal Industry

 I recently attended LexisNexis Canada’s Webinar “AI in the Legal Industry”  which explored the current adoption of AI in legal practice, the potential integration of AI in courts and legal institutions, and the legal implications of widespread use of generative AI tools like ChatGPT by the general public. Among other topics, the speakers addressed the impact of AI on fundamental legal principles, like access to justice, transparency, and the adversarial system, and the need for regulations and guidelines.

During the discussion, I found Abdi Aidid’s analogy comparing Large Language Models (LLMs) to an “eager but inexperienced associate” as a helpful way of understanding LLMs. 

The Eager but Inexperienced Associate 

First, Aidid reminds us that it's important to keep in mind that Large Language Models (LLMs) generate speech using syntactical structures and language patterns (31:10). As a result, they function probabilistically, trying to predict the most probable word or information based on the input context. Hence, LLMs do not ensure the provision of accurate or factual information; they only offer predictions of suitable information. When posed with a question, their responses may seem correct but can also contain fabricated or erroneous details, which is known as a "hallucination."

Aidid’s takeaway, was to understand these limitations and emphasize the importance of human supervision when interacting with Chat GPT and similar AI tools. Treating them “like an eager but inexperienced intern or associate” (at 32:30) in which LLM users must engage in a dialogical approach, providing feedback, and iterating with the LLM to get more reliable results (much like Daniel Schwarcz and Jonathan H. Choi’s Prompt & Confirm strategy).

Aidid's metaphor holds significance as it indicates the way forward for LLMs in practical applications, emphasizing a change in our professional relationship with technology. While other digital technologies may provide immediate and extractive outcomes, LLMs necessitate a more interactive and iterative approach. This shift in interaction calls for a redefined division of labor between technology and humans, requiring active human engagement and supervision to attain improved results.

To hear more, check out “AI in the Legal Industry; Exploring the Implications and Ethics of Rapid Innovation including Generative AI Tools.”

AI in the Legal Industry; Exploring the Implications and Ethics of Rapid Innovation including Generative AI Tools Abdi Aidid, Angela Swan, The Honourable Justice Joseph F. Kenkel, Jennifer Belz. LexisNexis Canada Webinar, 06.27.2023. 1:00:18