In this new era of artificial intelligence (AI) and machine learning, we’re starting to see its impact. AI-powered tools that are transforming numerous professions, including the legal industry. One such tool is ChatGPT, OpenAI’s language model that can generate human-like text based on the prompts it receives. Though it often generates very eloquent responses, it’s crucial to understand that its responses can be very, very inaccurate. Let’s dive into why this happens.
To comprehend the eloquence and inaccuracies of ChatGPT, we first need to understand how it operates. ChatGPT is a large language model that uses machine learning to generate text. It is trained on a vast dataset of text pulled from the internet. Through its training, it learns the patterns, structures, and nuances of human language. This lets it generate coherent — and very eloquent — responses to user prompts.
Eloquence in Writing
ChatGPT’s eloquence comes from its ability to learn and replicate the language patterns found in its training data. It’s able to formulate complex sentences, incorporate a wide vocabulary, and even mimic certain styles of writing. This feature is particularly noticeable when ChatGPT generates legal opinions. Legal texts often contain specialized vocabulary and formal structures which ChatGPT has learned and can reproduce. This eloquence has the added benefit — or pitfall — of inspiring confidence in the accuracy of ChatGPT’s responses. Unfortunately, this confidence is often unearned.
The Risk of Inaccuracy
Despite lawyers’ tendency to be verbose, legal opinions require more than just eloquent language. They require a deep understanding of statutes, rules, regulations, case law, legal doctrines, and more. They also require the ability to understand the context of a situation and to apply that understanding accordingly. ChatGPT, despite its sophistication, lacks several key abilities that limit its accuracy:
- Lack of Contextual Understanding. While ChatGPT works as a prompt-and-response system, it doesn’t truly understand the context or meaning of the information it’s processing. Despite how conversational the system is, its responses are based on patterns it has learned. It does not actually comprehend the subject matter it’s engaging with.
- No Real-time Data Update. ChatGPT’s information set was last updated in September 2021 and doesn’t update in real time. This means that it’s unaware of any recent changes or developments in the law. In the fast-evolving legal field, this can lead to outdated or inaccurate information. Consider the Supreme Court cases that have occurred just in this last year. Consider how different our legal landscape is from 2021. It’s vital that legal opinions include up-to-date information.
- No Personal Experiences or Opinions. ChatGPT doesn’t have personal experiences or opinions. It can’t weigh the merits of different legal arguments or make judgments based on experience or intuition. These are critical skills in the legal field. It’s often said that knowing your judge is as important as knowing the law. ChatGPT and other AI systems lack this sort of earned intuition.
- Inability to Guarantee Accuracy. ChatGPT generates responses based on probability. It chooses its words based on what it has seen most frequently or recently in its training data. This means it can’t guarantee the accuracy or appropriateness of its responses. It simply picks what it judges to most probably be the right response.
ChatGPT and the Practice of Law
To avoid the appearance of bias, I asked ChatGPT to ghost-write this article (which I then heavily revised). I wanted to see how it responds to questions about its own limitations. Moreover, I wanted to highlight to you, the readers, just how eloquent and convincing AI-generated content can be. The confidence it inspires with its eloquent responses is often unearned and occasionally quite dangerous. Even so, AI systems are progressing rapidly. ChatGPT’s latest model aced the bar exam, easily outpacing its previous version in only six short months. ChatGPT and other AI systems will soon become an integral part of the practice of law and many other professions.
AI has enormous potential in the legal field. Tools like ChatGPT can assist in drafting documents, conducting legal research, or providing very generalized legal advice. They promise to improve the efficiency of the practice of law while lowering costs for clients. Ideally, they could make the legal system more accessible for millions who might otherwise not be able to afford it. In that way, AI presents something of a watershed moment for the practice of law. Years ago, the advent of web-based legal research platforms made a ten-hour research process achievable in ten minutes. AI has the ability to further eliminate time-intensive legal processes, giving lawyers more time to focus on their clients.
While ChatGPT can write eloquent text and mimic legal opinions, it’s critical to remember that its responses can be inaccurate. AI systems are tools designed to assist — not to replace — human expertise, judgment, and oversight. It’s unlikely likely that AI will replace lawyers outright. Instead, lawyers may start to leverage AI to supplement their own practices for the benefit of their clients. Over time, that will simply become the expectation of both clients and the legal system at large.
We at Law 4 Small Business continue to be at the forefront of emerging technologies. Our goal is to use technology to provide the best, most effective legal services to our clients. Law 4 Small Business (L4SB). A little law now can save a lot later. A Slingshot company.