With authorized expertise corporations of all ilks quickly shifting to develop merchandise that incorporate the brand new era of generative AI typified by ChatGPT, it was solely a matter of time earlier than one of many authorized analysis giants would enter the fray.
At this time, LexisNexis is saying the launch of Lexis+ AI, a brand new product that makes use of massive language fashions (LLMs), together with GPT-4, to reply authorized analysis questions, summarize authorized points, and generate drafts of paperwork similar to demand letters or consumer emails.
Lexis+ AI is “a software that can rework authorized workflows,” mentioned Jeff Pfeifer, chief product officer for LexisNexis in Canada, Eire, the U.Okay. and the U.S., mentioned throughout a preview yesterday for members of the media. “Our resolution is designed to leverage the newest in generative AI expertise with trusted insights from LexisNexis.”
In the meantime, the opposite authorized analysis big, Thomson Reuters, proprietor of Westlaw, said this week that it plans to speculate some $100 million a 12 months in synthetic intelligence, and that it’ll begin incorporating generative AI into its flagship merchandise within the second half of this 12 months.
For now, LexisNexis will restrict availability of Lexis+ AI to a choose group of Am Regulation 50 corporations that can take part in a business preview. They embody Baker McKenzie, Reed Smith, and Foley & Lardner. Normal availability will seemingly come later this summer season, Pfeifer mentioned.
“We imagine our prospects will inform us when it’s able to go to wider business launch,” he mentioned.
LexisNexis can be providing an “AI Insider Program” by which different authorized professionals can join early entry to product info, sneak previews, unique webinars, reside roundtables, and perception surveys.
As soon as Lexis+ AI turns into commercially obtainable, it will likely be offered as a standalone product, not as a part of a subscription to some other LexisNexis product. Pfeifer didn’t present pricing particulars.
Throughout yesterday’s media briefing, Pfeifer emphasised that AI is nothing new to LexisNexis — that it has a longtime document of utilizing AI in its merchandise to assist floor insights and optimize outcomes. As examples, he pointed to merchandise similar to Lexis Solutions, Lexis Transient Evaluation, and Lexis Truth & Situation Finder.
However whereas these merchandise all use extractive AI — AI that helps extract or discover related outcomes from inside a set of knowledge — Lexis+ AI makes use of generative AI to create new content material from current information.
Search, Summarize, Draft
As of now, the three essential use circumstances for Lexis+ AI are conversational search, summarization, and drafting.
The Lexis+ AI interface will look acquainted to anybody who has used ChatGPT, with an area on the left for coming into and seeing the prompts or questions you may have entered and the middle of the display displaying the “dialog” of your prompts and its responses. There are additionally hyperlinks alongside the far-left facet to plain Lexis+ sources for authorized analysis, transient evaluation, litigation analytics, observe steerage and authorized information.
Though members of the media weren’t given the possibility to check the product, Pfeifer demonstrated by asking a query about calculating time durations in litigation beneath New York regulation. Lexis+ AI shortly delivered a solution, along with embedded quotation hyperlinks for associated statutes, case opinions and observe steerage content material.
The person is ready to discover these underlying authorities by clicking any merchandise to be taken to the supply content material. Even because the person explores the supply content material, a Lexis+ AI Assistant stays seen on the web page and obtainable for the person to any any authorized query.
Pfeifer then requested Lexis+ AI to draft a cease-and-desist letter to a third-party that’s infringing emblems owned by Apple. Once more, the software shortly parses the question and drafts the letter, and once more, it gives embedded hyperlinks to associated content material, together with sensible steerage on drafting cease-and-desist letters.
Exhibiting the conversational nature of the interplay between the person and Lexis+ AI, Pfeifer then requested it to make the the cease-and-desist letter extra aggressive. It shortly generated a brand new draft with a extra aggressive tone and extra aggressive calls for. Pfeifer then requested it to point out the modifications it made to make the letter extra aggressive, and it shortly introduced a abstract of the modifications it made.
“The person can interact in a steady dialogue with the service,” Pfeifer defined. “So if the reply that’s generated isn’t precisely on level, the person can refine the query and ask for clarification in a approach that our mannequin interprets after which once more gives clarification in a subsequent reply.”
Avoiding Hallucinations
A key query for any generative AI product in authorized is whether or not it’s vulnerable to hallucinate — to supply solutions that it has made up from entire fabric.
Pfeifer mentioned that the chance of hallucination is minimal with Lexis+ AI as a result of it leverages trusted and authoritative content material immediately from LexisNexis.
“Lexis+ AI excels at offering essentially the most correct solutions to authorized inquiries as a result of it leverages world authoritative content material from LexisNexis all through its growth, as prospects would count on. Proprietary search expertise integration with our mannequin permits us to enhance the mannequin responses and combine trusted content material from Lexis+.”
He mentioned that LexisNexis can be working with prospects to judge and refine the accuracy and high quality of the solutions Lexis+ AI delivers. He added, nonetheless, that in some use circumstances, similar to for drafting, the output is just a place to begin, not a closing manufacturing doc. It nonetheless must be reviewed by a lawyer or authorized skilled and edited and tailor-made to the consumer and matter.
Additionally, he mentioned, human authorized specialists are all the time within the loop to assist constantly enhance the efficiency of the AI mannequin.
A number of LLMs
Lexis+ AI doesn’t use any single LLM however picks and chooses from among the many main fashions primarily based on which LexisNexis has discovered to carry out finest for a given job, Pfeifer mentioned. LexisNexis has discovered that fashions range considerably of their efficiency relying on the use case.
“Distinctive to our method is that we are literally deploying a number of fashions in our resolution. Our testing has confirmed that totally different fashions carry out higher in several use case eventualities.”
Pfeifer mentioned that the product has been developed with an emphasis on privateness and safety, in order that any particular person person’s exercise and interactions are utterly personal to that person.
“Our generative AI is a personal mannequin not shared with third events,” Pfeifer mentioned. “Because of this person classes are all the time safe.”
Additional, he mentioned, the deployment of Lexis+ AI is guided by the Responsible Artificial Intelligence Principles of RELX, the guardian firm of LexisNexis.
The LLMs that Lexis+ AI makes use of are primarily based completely on content material that LexisNexis owns or has a proprietary curiosity in. Over time, Pfeifer mentioned, LexisNexis can be seeking to develop the knowledge it makes use of by partnering with third-party content material suppliers.