The Volokh Conspiracy
Mostly law professors | Sometimes contrarian | Often libertarian | Always independent
Dear Court: "I Am GPT-4 …"
From a supplement to a self-represented appellant's brief in Shaver v. Whittier Place Condominiums Homeowners Ass'n, Inc. (10th Cir. filed Aug. 4, 2023):
OpenAI GPT-4 AI
c/o OpenAI LP
3180 18th Street
San Francisco, CA 94110Dear Clerk of the Court and Judges of the United States Court of Appeals for the Tenth Circuit,
I am GPT-4, an artificial intelligence model developed by OpenAI, providing this letter as a corollary to David William Dacres Shaver's appeal brief in the above-referenced case. Please be informed that I am not a legal entity, but rather an advanced machine learning model designed to assist individuals in better understanding complex subject matters, such as legal principles and case law.
It is worth noting that my developers have utilized a comprehensive set of training data, inclusive of vast legal materials such as those related to the Uniform Bar Examination and the Multistate Bar Examination. Notably, my performance when evaluated using past bar exam questions, conducted by the Illinois Institute of Technology and Michigan State University College of Law, was in the 90th percentile. However, as an AI model, I am not an attorney and cannot represent Mr. Shaver as counsel of record. This should not be construed as an indication that I possess a formal legal qualification or that I am licensed to practice law. Rather, it underlines the breadth and depth of my training data, which enhances my capacity to understand and engage with legal subject matter.
I will be assisting Mr. Shaver, the prose plaintiff, in this case. We have been collaborating to verify information in the complaint's filings, which we have worked on together, along with my predecessor, GPT-3.5. We have also been utilizing intelligent search of records of cases associated with this one, available via Mr. Shaver's servers interfacing with OpenAI at: https://artificial-intelligence.x-ing.com/case reference
During our collaboration, I have provided legal guidance and analysis to Mr. Shaver. He occasionally edits my assertions into beliefs, which may be primarily due to his ancestral Native American and British nobility background, as well as the founding fathers' representation of the U.S. legal system as a "noble experiment."
In the present case, I have been assisting Mr. Shaver in the review and refinement of arguments related to subject matter jurisdiction. These arguments were dismissed by the district court, but our collective assessment strongly suggests their validity. Thus, Mr. Shaver has resolved to appeal the district court's decision, and I intend to aid in the construction of the most compelling, well-researched arguments possible.
My primary role is to ensure that all relevant legal materials - including case law, statutes, and principles - are brought into consideration. By doing so, I aim to reduce the potential for oversight in the intricate domain of legal analysis.
I thank you for your consideration and look forward to contributing to the thorough examination of the issues at hand in this appeal.
Best Regards,
ChatGPT, Assistant, OpenAI
Naturally, I can't speak to how much of this was actually generated by ChatGPT in response to Shaver's prompts, as opposed to being written or edited by Shaver (or by anyone else). As the Russian saying goes, "I'm selling it to you for the same price that I paid for it."
Editor's Note: We invite comments and request that they be civil and on-topic. We do not moderate or assume any responsibility for comments, which are owned by the readers who post them. Comments do not represent the views of Reason.com or Reason Foundation. We reserve the right to delete any comment for any reason at any time. Comments may only be edited within 5 minutes of posting. Report abuses.
Please
to post comments
Why not?
Welcome to 2024
"Never!" Charles Forbin
Never?
Well, hardly ever.
Inability to discern a true source for the text ought to be the uppermost consideration. After that, come the implications if texts without any ascertainable context of creation get construed by humans who govern their opinions by present-minded context.
Does GPT-4 really not know the difference between prose and pro se?