Take a second and navigate to chat.openai.com, the URL belonging to the now infamous ChatGPT Artificial Intelligence Large Language Model (AI LLM). The dialogue box at the bottom prompts you to send a message to ChatGPT, as if it were a flesh and bone individual on the other end. Type a question or request out and you’re almost instantaneously shown a response that, while it doesn’t quite feel human, feels far from the simple accuracy that IBM’s Watson gave on its run to a Jeopardy championship. That is the nature of modern AI; not simple one-word answers but eloquent, often well thought out responses. But what impact might this technological development have on the legal
profession? CSU Law’s Cyber Law Society (CLS) recently explored one of those potential impacts at their “AI Writes the Law” event.
Armed only with an AI LLM, either OpenAI’s ChatGPT, Google’s Gemini, or Microsoft’s CoPilot, individuals were asked to create a law that prohibited the use of vehicles in the park. There were no rules, save that their submissions, the laws they formulated with the AI LLMs, must consist only of the AI LLM’s responses. By exploring the abilities of each of these different AI LLMs, the group came away with some very interesting conclusions.
Most interestingly, each AI LLM had a different response when prompted to write a law. While ChatGPT dove right in and produced a rough law, Gemini initially stated that it could not write a law and suggested the user should consult an attorney. After explaining that the law would be for purely non-regulatory and non-legislative purposes, Gemini conceded. After much refining, all three of the AI LLMs produced well written laws prohibiting vehicles in the park, but one stood out from the rest.
Microsoft’s CoPilot, fueled by Dylan Ramsey’s input, created far and away the best law, and it wasn’t particularly close. Each AI LLM did well in defining what a vehicle was for the purposes of the law, but CoPilot went beyond by taking into account nearly any type of vehicle and scenario possible. CoPilot’s law included schedules for fines and noted precisely where those fines would be going. CoPilot even required transparency in the fund allocation process as part of the statutory text.
Each AI LLM produced a functional and detailed law, but none of these laws would be possible if not for the well thought out inquiries by the users. All of these AI LLMs draw from a massive wealth of information, and to make the best use of that information, any user must be able to work with the AI rather than simply use it as a tool.
In the end, regardless of whether you think AI is coming for your job or that it’s not worth losing sleep over, one thing we should all agree on is that it has great potential to make the practice of law more efficient and allow its practitioners to draw on a wider array of knowledge than ever before.
Comments