A previous edition of this article stated that ChatGPT could be used to summarise judgments. We have since amended the article to reflect that it is best used (with due diligence) on older cases.
BEN CHESTER CHEONG SHARES 3 IMMEDIATE USES FOR CHATGPT
Mr Ben Chester Cheong, a lecturer from the Singapore University of Social Sciences
BY ASHUTOSH RAVIKRISHNAN
ChatGPT is nearly impossible to escape these days: the chatbot has made its way into everyday parlance thanks to the speed of its adoption and its impressive accuracy. And hot on its heels is Harvey, Allen & Overy’s artificial intelligence chatbot, which lawyers can use to help draft contracts, according to the Financial Times.
To help members make sense of the possible implications of these chatbots for the law, we spoke to Mr Ben Chester Cheong, a law lecturer from the Singapore University of Social Sciences. He points out that lawyers should think outside the box when it comes to using ChatGPT.
“Legal drafting is the first process that comes to mind, and with good reason,” he explains. But while generative technology has improved its legal drafting skills by leaps and bounds, there are still inaccuracies and errors, which may make it a less productive exercise than assumed.
But there are other purposes that SAL members can consider using ChatGPT for, says Ben.
1) Creating marketing and business development collaterals
“An associate who needs to prepare a marketing or business development deck could turn to ChatGPT for a proposed outline,” explains Ben.
We put his theory to the test and true enough, a marketing outline that covered the bases was generated in a few seconds:
2) Making quick sense of legal concepts
“Students see ChatGPT as a tool to help them understand legal concepts. For example, some students may struggle with certain equity and trusts concepts like unjust enrichment. I think ChatGPT does a good job in summing these up and helping students make sense of them.”
Once again, we turned to the chatbot to illustrate his point.
Here’s a standard definition of unjust enrichment from Google, which pulled the data from Cornell University:
Unjust enrichment occurs when Party A confers a benefit upon Party B without Party A receiving the proper restitution required by law. This typically occurs in a contractual agreement when Party A fulfills his/her part of the agreement and Party B does not fulfill his/her part of the agreement.
The definition above reads like legalese, which does little to help a layperson understand the law. On the other hand, ChatGPT’s explanation comes with an example, which goes a long way in helping students and laypeople understand the law:
Sure! Unjust enrichment is a legal concept that refers to situations where one person has received a benefit at the expense of another person, without any valid reason or legal justification. In such cases, the law may require the person who received the benefit to return it or compensate the other person for its value.
Let's say that Alice hires Bob to paint her house for $5,000. Bob starts painting the house and buys all the necessary supplies, including paint and brushes, for $1,000. However, before Bob can finish painting the house, Alice decides to terminate the contract and asks Bob to stop working. Bob has already spent $1,000 on supplies and has worked for a few days, but Alice refuses to pay him anything for his time or the supplies he purchased.
In this scenario, Alice has been unjustly enriched at Bob's expense. Bob provided a service by starting to paint Alice's house, and he also incurred expenses by purchasing supplies. Alice received the benefit of Bob's work and the supplies he purchased, but she did not compensate him for either. This would be an example of unjust enrichment, as Alice has been enriched without any legal justification and at Bob's expense. Bob would have a legal claim to recover the reasonable value of his services and expenses from Alice.
The basic idea is that when one person receives a benefit at the expense of another person without a valid reason, the law may require the person who received the benefit to make restitution to the other person.
See the difference?
3) Processing data quickly (not all data though)
Judgments can run into tens of pages, and with good reason. But when you’re pressed for time, you may be tempted to use ChatGPT to quickly summarise a judgment. However, be warned that for newer judgments, ChatGPT is prone to "hallucinate", as is the case below (the case has nothing to do with PVC resin, although that only came to our attention after this article was published):
In the case of older judgments, users have to be fairly detailed in their prompts. "Only with sufficient context would it be able to then summarise content," he says. "But from what I understand, A&O's Harvey is based on GPT-4. This has been trained specifically for the legal market, which means they have fed it with enough legal materials, cases and precedents for it to churn out something accurate, even for judgments. But as to how Harvey is, it seems only A&O is privy to it for now."
That being said, lawyers can still use ChatGPT to summarise non-legal content for the time being.
As we end our conversation, Ben brings up a software called Contract Express that he used around five years ago when he was in practice. “You need to have some technical knowledge in order to help the software learn to churn out first-cut documents. But in just five years, generative AI, the tech behind such software, has improved so much that people don’t need technical knowledge to use ChatGPT." He believes that further advances in generative AI will soon make tech like ChatGPT indispensable to the legal profession. “It’ll be like what Microsoft Word is today—part and parcel of practice.”