ChatGPT is a new and wildly popular generative artificial intelligence (AI) tool from OpenAI.[i] Users provide textual instructions to the tool, such as “Explain quantum computing in simple terms,” and the tool will generate a response. Based on people’s reactions so far, the responses are uncannily realistic.
The release of ChatGPT has stoked speculation about all the different fields it could potentially disrupt. Educators are worried that students could use the tool to write school essays. Programmers are worried that the tool may allow anyone to write functional code. Even Google is worried about the tool replacing the need for conventional search engines.
What about the legal industry?
It’s possible that generative AI could disrupt the legal industry, but it’s difficult to say for certain without knowing more about the specific applications of the technology in the field. Generative AI is a type of artificial intelligence that is capable of generating new content based on a set of input data. This could potentially be used in the legal industry to automate the creation of legal documents or to assist lawyers in researching and analyzing complex legal issues. However, it’s important to note that the adoption of any new technology can be a complex process, and it may take some time for generative AI to be fully integrated into the legal industry.
You may be surprised to learn that the above paragraph was generated entirely by ChatGPT in response to the input, “Will generative AI disrupt the legal industry?” It is remarkable that, less than 50 years after the first word processing programs were introduced and ran on eight-bit microprocessors, computers are now able to write realistic, human-sounding text on their own.
But despite looking and sounding realistic, the responses provided by ChatGPT have some inherent shortcomings that are problematic in the legal field.
1. The responses are not always accurate.
In the legal profession, accuracy is paramount. Providing inaccurate legal advice harms clients and places the lawyer at risk of liability for malpractice.
The responses provided by ChatGPT have proven sometimes inaccurate.[ii] OpenAI includes several prominent disclaimers on its website indicating as much. As one example, OpenAI writes in an FAQ answer that the responses “may be inaccurate, untruthful, and otherwise misleading at times.”[iii]
This inaccuracy results from the inherent structure and operation of ChatGPT. The tool uses sophisticated deep learning techniques and artificial neural networks to generate complete responses to a textual input by predicting the most probable next word in a sequence based on the input and preceding text.[iv] In other words, ChatGPT is simply a language prediction model. It lacks the ability to truly comprehend the meaning behind human language and conversation.[v] Furthermore, ChatGPT may be trained on data that itself is not entirely correct. The age-old maxim holds true – garbage in, garbage out. Consequently, no one can afford to blindly trust the results of ChatGPT to provide accurate legal advice.
2. The responses do not provide attribution.
Providing a source in support of a legal proposition is a fundamental aspect of legal writing. But ChatGPT currently has no way to provide attribution. ChatGPT predicts each word in a response in succession based on a “black box” neural network methodology rather than taking an entire concept from a single, definable source. Thus, it is not possible for ChatGPT to provide proper attribution. Nonetheless, this may be an area where ChatGPT or other tools could improve in the future.
3. The responses do not apply well to complex or novel issues.
The inherent structure and operation of ChatGPT also does not lend itself well to providing responses to inputs that make complex or novel requests. Again, the tool is based on replicating existing language and has no actual understanding of the language or issues. Testing a few inputs asking complex legal questions, such as, “Why is U.S. Patent No. #,###,### valid over Alice?” predictably yielded useless responses.
The shortcomings identified above necessarily preclude ChatGPT from outright replacing the function of lawyers. However, ChatGPT still has significant potential in revolutionizing how lawyers work. Previously, computers and typewriters enhanced the productivity of lawyers by enabling them to type briefs rather than write the briefs by hand. Legal research tools like WestLaw and Lexis also enhanced the productivity of lawyers by enabling them to search electronically a wide universe of case law rather than manually reading through volumes of books. ChatGPT may similarly enhance the productivity of lawyers. Armed and guided by the strategic thinking, organization, and planning of a lawyer, ChatGPT is likely to assist lawyers in tasks such as communicating with clients, creating legal documents, and performing legal research, but cannot replace actual competent legal analysis and advice.