Lawyers who expect generative artificial intelligence to significantly impact the practice of law see some of the greatest potential in legal research and writing.
In August, a LexisNexis Legal & Professional study of nearly 8,000 lawyers, law students and consumers in the United States and three other countries found 65% of these professionals believe generative AI tools could assist them in researching matters. Meanwhile, 56% believe the tools could help them draft documents.
Ed Walters, CEO and co-founder of online legal research software company Fastcase, is thrilled by the prospect of making legal research more efficient. Traditionally, researchers would have to come up with keywords and add them to mathematical strings of language that they plug into a search engine. They would then receive a long list of documents and often spend hours reviewing each result to see if it yields their answer.
“This way of doing research is clumsy and slow and often leads to wrong answers,” says Walters, who also became the chief strategy officer of vLex after the legal technology company merged with Fastcase in April.
But now Walters says lawyers could ask AI tools to pull together relevant documents, read them and instantaneously synthesize results for them. In October, his company released a new version of Vincent AI, a legal research assistant that, among other features, allows lawyers to get answers to legal questions with citations and links from verified sources.
He expects that with the advent of GPT-4 and other large language models that interact with vast quantities of text, more of these tools with similar and even expanded capabilities are just on the horizon.
“This generation of tools doesn’t really solve research tasks, but I think they point to a new generation of tools right after this that will, instead of creating statistically likely answers, understand there’s a question being asked and synthesize text to answer the question,” Walters says.
June Hsiao Liebert, the president of the American Association of Law Libraries, agrees that large language models show a lot of promise for improving legal research and writing. She says they are already useful for some steps involved in these tasks, including drafting and editing documents.
“Given how they work, tools utilizing [large language models] can be useful in modeling what the ‘average’ or ‘typical’ solution or document might look like for a particular question, since it essentially predicts words based on what it has been trained on,” says Liebert, the director of information services at the law firm O’Melveny & Myers. “It can also give users different answers depending on the point of view, set of facts or purpose. This ability to model different options can be extremely helpful and time-saving.”
Liebert encourages lawyers who use generative AI to double-check their work but says legal tech companies are making progress in addressing problems with accuracy in available tools.
She anticipates they will continue to investigate and devise solutions for other issues, including authenticity and bias.
“As these tools improve and we become better at using them, [large language models] and other AIs of the future will become just another tool that is available to everyone in the legal industry, not unlike the spell-check tool or even calculators,” Liebert says.
Locked in: Criminal justice startups tap into generative AI’s early promise
Bridging the Gap: Lawyers trying to increase access to justice see promise in generative AI
Helping Hand: Generative AI already is making an impact on legal research and writing
Always on: Will generative AI alleviate burnout or make lawyers more miserable?
Head of the Class: Law schools consider post-ChatGPT coursework
Age of e-Discovery: Generative AI could revolutionize e-discovery—but buyer beware
Rewiring Entry: How AI could blur the borders of immigration law
This story was originally published in the February-March 2024 issue of the ABA Journal.
