The complex field of artificial intelligence (AI) continues to push the frontiers of legal practice, ushering in new challenges and opportunities. In the early days of electronic legal research, mastering Boolean logic was the ticket to success. Skilled attorneys could combine legal acumen and mastery of search syntax to quickly pinpoint relevant case law.
The development of AI, however, saw the rise of natural language processing models, evening the playing field and reducing the competitive edge gained through Boolean mastery. As we enter the era of generative AI, the art of crafting effective prompts has emerged as the new determinant of success. Aspiring legal professionals must now master the art of framing AI prompts, similar to their predecessors’ mastery of Boolean logic. However, achieving this proficiency is far from simple, given the still-evolving landscape of AI tools and applications.
Case in point, AI precision isn’t yet sophistication enough to be fully implemented in the legal industry. And to engineer an effective prompt, one must understand what the AI is looking for – a challenge considering the complexities involved.
There’s a growing need for legal professionals to understand how Large Language Models (LLMs) react to prompts—what it uses, ignores, and the work it does based on the prompts. The spotlight nowadays is on developing tests to determine these. As Jeremy Pickens, Head of Applied Science at Redgrave Data, puts it, the relationship between the AI’s understanding and our prompts is essentially a token problem — getting the desired output without being burdened by an excess of tokens.
Unchallenged, these factors pose a risk to the widespread adoption of AI in law. Even with the current popular usage of AI for summarization, the biases within the AI system might affect the output critically. Therefore, the ball is in the court of scientists and developers, who must develop robust, clear, predictable LLMs that satisfy legal requirements, and legal professionals who must swiftly adapt to leverage them.
Further, the lawyer of tomorrow might be the one who understands how to use AI most effectively, but what that entails remains to be seen. Indeed, the field of AI in law is still in its infancy, and the path to AI literacy is not clearly charted (Patrice, Above the Law).