- Beware of Defamation Liability for Reposting Material on Social Media
- The Experiment Failed: AI Isn’t Yet Ready to Be a Good Document Summarizer
- When Trademark Rights Collide with Free Speech, Which Wins?
- Fair Use Protection Against Copyright Infringement Claims Just Got Smaller
- Will Copyright Law Wipe Out Generative AIs Such As ChatGPT?
Latest Blog Posts
Can ChatGPT, as it currently exists, make lawyers more efficient and, thus, less expensive?
As you probably know, it’s a generative artificial intelligence (AI) chatbot. It can crank out authoritative-sounding, well-written prose.
It also sometimes says crazy things. In a recent experiment using ChatGPT through the Microsoft Bing search engine, it expressed desires to steal nuclear codes, start a pandemic, and be alive.
I’ve experimented with it and read other lawyers’ reviews. It isn’t ready for legal prime time. But the day will come, perhaps soon, when it or another AI will be a critical companion to practicing law cost-effectively.
What lawyers do depends upon their specialties. Still, most kinds of legal practice involve drafting correspondence, performing legal research, writing memos and briefs, drafting legal instruments such as contracts, doing fact research, brainstorming legal theories, and lots of editing.
Except for editing, all those tasks require a lawyer to know the applicable law, find the relevant facts, and apply the law to those facts. ChatGPT does poorly in those areas.
ChatGPT is often confidently wrong (although there is AI technology that might largely fix this in future chatbots). If you ask it to state the law in a particular area, it will give you an authoritative-sounding answer. But ChatGPT’s answers sometimes omit key elements, misstate fundamental principles, or cite authority inaccurately, such as identifying the wrong court.
That means the lawyer must know the law to know when ChatGPT messes up. Thus, not much time is saved. Still, it might be good for discovering some statutes, regulations, or cases the lawyer didn’t know about, but the lawyer will still have to verify those authorities.
Keep in mind that experienced lawyers rarely start a project from scratch. Whether it’s writing a brief, memo, contract, or corporate document, experienced lawyers often start with the closest thing they have created in the past and then adapt it to the client’s need, perhaps also drawing on other similar past work. That makes the lawyer efficient.
Then there is gathering the facts. ChatGPT isn’t the tool for that. Much of the needed information will come from the client. Perhaps ChatGPT could help find relevant online information, but, to be usable, you will need to be able to link individual facts to trustworthy sources.
Next, lawyers must worry about protecting the attorney-client privilege when using services such as ChatGPT. Information you input into a generative AI might be incorporated into the output that someone else gets from the AI (the extent to which this happens is unclear). Also, you might be forced in litigation to divulge any information you input into an AI because that inputting might constitute disclosing information to a third party in a way that destroys the privilege. This issue is fixable if future generative AIs have the right confidentiality policies.
I don’t think we will reach the point where a human lawyer’s knowledge of and experience with the law and a human’s factual analytical ability will be entirely supplanted by AI. But we will reach the point, maybe soon, where the tool becomes accurate enough that lawyers will be forced to work with it to be price competitive.
Also, it would be an unfair comparison to expect perfection from a law AI. Many lawyers produce way-too-brief, incomplete, or shoddy work products, even lawyers at big-name law firms.
But those human errors tend not to entirely defeat the purpose of the legal work. An AI could make a mistake that is small in the computer realm but large in human impact, such as adding extra zeros in stating an amount of money. Can law AI develop sufficiently to preclude such big mistakes? Until it can do so, law AI won’t save the need for a lawyer specializing in the same field to review and validate the AI output.
AI won’t invade every aspect of legal practice in a big bang. Instead, lawyers will slowly add it to their toolkit, one AI tool at a time.
Lawyers already use AI tools for some purposes, such as grammar-checking and editing services, natural-language legal research functionality in legal databases, information research services built for specific legal specialties, such as for trademark-clearance research, and algorithms for sorting through electronic documents in discovery in litigation.
In fact, some legal vendors offer AI services for drafting, reviewing, and suggesting alternative language for contracts. I don’t know if those services are sufficiently strong and accurate yet.
Even though there are claims of ChatGPT passing law school and practice bar exams, ChatGPT has a long way to go before we should give it a license to practice law.
Written on February 22, 2023
by John B. Farmer
© 2023 Leading-Edge Law Group, PLC. All rights reserved.