Today’s headlines from newspapers such as The Hill, The Register, and Bloomberg were nothing short of expected. We expressly known for some time the inadequacy of consumer-facing generative AI tools, such as ChatGPT and related language models, for legal work.
In fact, these limitations have led to numerous incidents in recent years. Lawyers have been sanctioned, with some seeing their licenses suspended due to the ill-advised use of these AI tools. Even high-profile figures, such as former President Trump’s lawyer, fell prey to the pitfalls of these inaccurate platforms.
The media buzz around these AI models stems from a recent Stanford study that critically looked at the frequency of legal errors in gen AI outputs. However, this study essentially brings nothing new to the table, merely quantifying the issues already widely recognized.
In reality, AI development is already evolving to address these shortcomings, with GPT4 and Llama3 on the horizon. Rather than focusing on generic AI tools, industry insiders are honing in on trusted datasets and legal-minded guardrails to hone accuracy and improve the quality of legal applications.
Nonetheless, these headlines and studies hold some value for the broader public, particularly those endeavoring to navigate legal questions without legal counsel. The Stanford study highlights the risk of relying on these AI bots for legal guidance, particularly showing a higher prevalence of misinformation in lower court case law, typically relied upon by individuals seeking justice autonomously.
For full article, see post on Above the Law.