By Matt Wetherington | January 30, 2023 at 09:01 AM
Originally published in the ALM Daily Report
The Rise of ChatGPT
ChatGPT is an advanced artificial intelligence program that promises to revolutionize the legal profession. Developed by OpenAI, this cutting-edge technology utilizes natural language processing models to generate human-like text, solve complex coding issues and provide valuable insights into large data sets.
The potential applications of this technology are limitless and have the potential to greatly improve the efficiency and accuracy of legal work, while also creating new problems that require thoughtful solutions.
It’s important to note that ChatGPT does not plagiarize in the traditional sense of the word. The output produced by ChatGPT is unique and different from the data it has been trained on, meaning that the same question will rarely generate the same response. The output is a combination of the knowledge from the data it has been trained on and the input provided to it.
OpenAI has made the model available to the public, and anyone can access and use the technology at no cost for the time being. This presents a great opportunity for lawyers to familiarize themselves with the technology that may soon play a significant role in their profession.
ChatGPT has already demonstrated its capabilities by outscoring human beings on the evidence and torts sections of the bar exam. But don’t expect it to start drafting filing-ready briefs yet. Although ChatGPT was able to draft many of the sentences in this article, it was unable to make a polished one, even after submitting substantial revisions.
Wrestling and Plea Bargains
Every practice area can benefit from a legal chatbot that can spend unlimited time answering client questions and explaining complex legal questions in familiar terms, like using pro wrestling to explain a plea bargain:
Just as wrestlers have a predetermined outcome for their match, a plea deal is a pre-determined outcome for a criminal case. In the same way that wrestlers work together to create a match that advances both of their storylines, a prosecutor and a defendant work together to reach a mutually beneficial agreement. And just like a face wrestler may choose to let the heel wrestler win to set up a rematch with a different outcome, a prosecutor may agree to a plea deal in order to secure a conviction for a lesser crime or set the defendant up for a rematch with more serious charges later.
ChatGPT, like any AI technology, is only as good as the data it is trained on. ChatGPT was trained on trustworthy and untrustworthy data sources. It also reaches wildly incorrect conclusions sometimes, particularly related to math. It is crucial to understand the sources of AI data and its limitations ensure that the final product produced is accurate, unbiased, and relevant to the task at hand.
Use Your Data to Overcome Problems
One way to overcome data coverage problems is by training the models with your own data. This can help to ensure that the results are specific to the task at hand and more accurate. For example, I submitted all briefing and orders I could find from seven civil cases decided by the Georgia Supreme Court in 2022. I then asked several questions about the strength of the briefing and which party should prevail. ChatGPT correctly identified the disposition of six out of seven cases.
It refused to assess the seventh case and instead encouraged me to hire legal counsel. I then uploaded the entire record and all briefing from a currently pending appeal and asked for an assessment. Thankfully, it determined we should prevail. However, adding personal information to ChatGPT raises data privacy and security concerns.
Lawyers have a duty to protect our clients’ confidential information and ensure that ChatGPT is used in a way that does not compromise this duty. Thus, as AI makes our work easier and faster, we have a responsibility to educate our clients about the implications of using AI in legal matters.
We should be transparent and explain the potential benefits and limitations of using ChatGPT in our work, and ensure that our clients are informed and give their consent before using the technology on their behalf.
ChatGPT Not Absolute
Furthermore, we should also be aware of the limitations of ChatGPT’s output and make sure that the output is not taken as an absolute truth but rather as tool to assist in our research and analysis. We should be mindful of the fact that AI is not infallible and that human judgment is still required, at least for now.
Finally, as ChatGPT becomes more prevalent, legal issues related to AI’s accountability, liability and transparency will need to be addressed. We must consider how to ensure that AI-generated outputs are reliable, fair and trustworthy.
We must also consider issues of bias and fairness in the use of ChatGPT. As the model is trained on a specific set of data, it may inadvertently reproduce any biases present in that data. It’s our responsibility to ensure that the use of ChatGPT does not perpetuate discrimination implicit in its underlying data sources.
In conclusion, ChatGPT is a powerful tool that will usher in dramatic changes to the legal profession. However, it is important to understand its capabilities and limitations, and use it responsibly. And, even if you choose not to use it for legal work, it does an excellent job writing bedtime stories and whipping up delicious recipes.
Attorney Matt Wetherington is the founding partner of the Wetherington Law Firm in Atlanta. His practice focuses on high-stakes cases involving personal injury, wrongful death and class actions.