AAJ Trial Magazine Feature: Jumpstart Your AI Journey
Contents
The potential uses for generative Al and large language models run the gamut. Start exploring how to implement this technology in your practice.
Click here to download this article as PDF | Reprinted with permission of Trial® (June 2024)
In November 2022, OpenAI launched ChatGPT, a generative-Al large language model (LLM) chatbot.1 It was a revolutionary technology with the capability to write effective computer code, compose music, draft research papers, create comprehensive business plans, generate Excel formulas, and even give relationship advice. By January 2023, ChatGPT had over 100 million users.2
This technology's immense potential also brought panic. Its ability to believably mimic student-created work sent shockwaves through the education field, resulting in calls to ban students from using the technology.3 Last year's television and film writers strike was driven in part by substantiated fears that studios would increasingly turn to Al to write scripts.4 And financial powerhouses such as Bank of America, Goldman Sachs, and JPMorgan restricted employee access to ChatGPT.5
The legal field is no exception. Some lawyers have faced consequences for using ChatGPT for research and submitting court filings that referenced cases that didn't exist.6 Despite the early trepidations, a recent study found that 73% of lawyers plan to use generative AI at work within the next year.7
At this point, we all know that AI is here to stay. Many will be able to leverage the technology to improve their research and writing-and ultimately streamline litigation. But this technology is new, and ethical guidelines are still being developed. Here's a practical guide to using generative AI effectively and responsibly.
What Is Generative Al?
When asked to define "in layman's terms" what generative AI is, ChatGPT explained it as "computer programs or systems that can create new content or data, such as text, images, or music, without being explicitly programmed for each specific output.”8
Generative AI differs from traditional AI (also known as "weak Al" or "narrow Al") in that the traditional models are designed to perform specific tasks or solve specific problems within predefined boundaries, without the ability to adapt or learn from new data beyond their initial programming or to create new content.9
In other words, generative AI is unlike other computer systems because it can instantly sift through enormous amounts of data, learn from that data, and then make decisions based on what it has learned while factoring in previous training and user interactions. Essentially, generative Al processes information in a manner comparable to humans, except exponentially faster and with the ability to process larger amounts of information all at once.
Checks and Balances
As we enter the age of AI, lawyers need to establish checks and balances around their use of this technology. With the great power Al can provide also comes a great responsibility to use it ethically. It's important to establish a framework of considerations to keep in mind, especially as generative Al technology continues to evolve. This framework should include:
- Accountability. Understand Al's capabilities and limitations and consider anything it produces for you or your practice as an extension of yourself. Al's mistake is your mistake.
- Transparency. As a leader of your firm, it's essential that you set clear guidelines for what Al can and can't be used for, in conjunction with open and honest discussions of the topic with your team. For example, this can be in the form of an AI handbook that outlines specifically when the technology can be used (such as for summarizing large amounts of anonymized data or information into a more digestible format) and when it can't be used (such as for drafting entire complaints or motions).
- Data privacy and confidentiality. You must educate yourself and your staff on the privacy and confidentiality risks associated with AI and how to minimize those risks. Specifically, this means teaching your team how to anonymize potentially sensitive data before uploading it into an LLM.
- Professional responsibility. Attorneys have a responsibility to engage in constant study and education on both the law and its practice. This includes the ever-evolving role of technology. Generative Al is meant to supplement-not replace—a lawyer's knowledge.
What Al Can't (or Shouldn't) Do
Artificial intelligence is far from perfect. Particularly, Al occasionally "hallucinates," confidently producing incorrect or nonsensical responses.10 Therefore, it's imperative that lawyers review everything that Al tools produce. It is best practice not to take AI at face value. Instead, use it as a supplemental tool to summarize data or documents, organize your thoughts, or even take the first attempt at drafting content. Al should never be used to write a motion that is then filed without any review, for example. Artificial intelligence should never act as a replacement for the human reasoning and judgment that will always remain an essential component of legal practice.
Data security is also a paramount concern. LLMs, by default, keep a running record of everything you input to learn and improve. It's also possible for generative AI software to be hacked. Therefore, any confidential case or client data should be redacted or anonymized before inputting it into an LLM.
Also keep in mind that AI has been found to elicit gender and racial biases in the content it produces.11 This reinforces the need to always make sure a human set of eyes reviews everything produced by AI-from a filing to a client email-before it leaves your office.
How Can Generative Al Be Used?
Generative Al can help you and your firm litigate and operate more efficiently. Here are some ways you can use this technology.
Legal research and summarization. Generative Al is a potential gamechanger when it comes to legal research and analysis. Instead of spending countless hours sifting through documents, medical records, peer-reviewed studies, and other research materials, lawyers and paralegals can drop these documents into an LLM and prompt the software tool to present a summary based on their desired criteria.12 (Remember: It is crucial to redact or anonymize any confidential or client data before using an LLM tool to protect privacy and confidentiality.)
For example, if a client is alleging a back injury caused by a defendant's negligence, instead of combing through hundreds of pages of medical records to put together a chronology to support causation, plug the anonymized records into the LLM and prompt it to create the chronology for you-which it can do in seconds.
Now, consider this capability on a larger scale, say in multidistrict litigation with thousands or tens of thousands of claimants. The ability to summarize a lifetime of medical history and identify causation in seconds for thousands of claimants would save time and resources. For example, since generative Al tools can process large amounts of data and identify patterns based on what the user asks of them, you could enter someone's anonymized medical history and ask the tool to pull out every time treatment was provided for a specific injury.
As a research assistant, generative Al can help lawyers glean valuable insights from overwhelming amounts of data-it can help identify the information most pertinent to the litigation, enhance efficiency, and better inform decisions. By using AI, lawyers can avoid getting lost in the information overload that can often hinder traditional research methods. Sometimes "more" can mean more complexity and confusion, but AI can point legal professionals toward the critical data they need to avoid falling down data rabbit holes.
Intake. Generative Al can streamline client intake for firms, saving time and money. It can automate the initial interaction with potential clients and help create a smooth and efficient process from the beginning. And it can swiftly gather and analyze relevant information for personalized client communications.
Law firms can provide clients with 24/7, real-time updates on their cases through LLM chatbot solutions embedded within their websites. Additionally, Al-powered case evaluation software can — using personalized, preset criteria — review cases for intake and move relevant claims forward for attorney review. Used correctly, AI empowers law firms to provide a higher standard of service and foster stronger client relationships.
However, it's essential for firms to find the right software partner that not only understands AI but also understands the unique complexities of the plaintiff bar. An Al chatbot solution provider that works primarily with online clothing stores, for example, likely won't be able to properly tailor their solution to uniquely suit your firm.
Deposition assistance. Depositions can be extremely time-consuming, and it's often difficult to parse transcripts and pull out the revelations that can make a case. Artificial
intelligence can help ease the burden of sorting through lengthy depositions. While live transcription is nothing new, transcription tools integrated with generative AI can analyze the transcripts and, in real time, automatically separate important responses. Additionally, AI can immediately flag relevant documents and exhibits based on responses from the deponents and can note pauses, unusual body language, and discomfort during live depositions.
Jury selection. In the context of assembling profiles about potential jurors, Al uses advanced data analysis techniques to assess potential jurors' backgrounds, social media activity, and public records. It can identify patterns, biases, and attitudes that might impact the case. This information helps you make informed decisions during jury selection and can increase the likelihood of selecting an unbiased jury.
Damages analysis. Generative AI can analyze large volumes of legal precedents, case histories, and relevant data to estimate potential damages in a more accurate and consistent manner than human analysis. It can consider a range of factors, such as past verdicts, economic indicators, and the specific circumstances of the case to provide a data-driven assessment of potential damages. This aids lawyers in determining the appropriate amount for settlement offers or in presenting compelling damages arguments in court.
Transactional law. Lawyers can use generative AI to automate essential but tedious contract-related functions, allowing more time for strategizing and negotiating. Specifically, generative Al can ensure lengthy contracts are uniformly formatted within seconds, streamline version control to ensure changes and revisions are accurately tracked, review and assess boilerplate clauses, and instantly proofread documents for typos and grammatical
Virtual legal assistants. Large language models can carry out many of the same tasks as a legal assistant but, in some cases, with greater efficiency. These tools are great notetakers, scheduling assistants, and inbox organizers. They can help organize your thoughts into cohesive emails to clients, opposing counsel, and insurance companies.
Of course, using Al as a personal legal assistant takes time and effort. Just like a human legal assistant, there's a learning curve for you and the LLM. You need to learn how to correctly prompt it to produce the desired response. It needs time to improve its functionality based on repeated interaction. And you should always double check everything AI produces for tone and accuracy.
Generative Al has taken the world by storm and ushered in a transformative era for legal professionals. Its potential to enhance overall efficiency is undeniable.
But it's also essential to recognize that AI is fallible and must be used judiciously.
Lawyers must exercise caution, ensuring AI supplements rather than replaces the human decision-making necessary to litigate clients' cases effectively and ethically. It's imperative to maintain data security and address potential biases in AI-generated content. If used responsibly, Al can enhance your speed and effectiveness as a trial lawyer, providing a strategic advantage in the fight to deliver justice on behalf of clients.
The views expressed in this article are the author's and do not constitute an endorsement of any product or service by Trial or AAJ.
1 Bernard Marr, A Short History of ChatGPT: How We Got to Where We Are Today, Forbes (May 19, 2023), https://www.forbes.com/sites/bernardmarr/2023/05/19/a-short-history-of-chatgpt-how-we-got-to-where-we-are-today/?sh=5187e78a674f
2 Martine Paris, ChatGPT Hits 100 Million Users, Google Invests in AI Bot and CatGPT Goes Viral, Forbes, Feb. 3, 2023, https://www.forbes.com/sites/martineparis/2023/02/03/chatgpt-hits-100-million-microsoft-unleashes-ai-bots-and-catgpt-goes-viral/
3 Kalhan Rosenblatt, ChatGPT Banned From New York City Public Schools' Devices and Networks, NBC News, Jan. 5, 2023, https://www.nbcnews.com/tech/tech-news/new-york-city-public-schools-ban-chatgpt-devices-networks-rcna64446
4 Andrew Dalton & The Associated Press, Writers Strike: Why A.I. Is Such a Hot-Button Issue in Hollywood's Labor Battle With SAG-AFTRA, Fortune, July 24, 2023, https://fortune.com/2023/07/24/sag-aftra-writers-strike-explained-artificial-intelligence/
5 Brian Bushard, Workers' ChatGPT Use Restricted at More Banks — including Goldman, Citigroup, Forbes, Feb. 24, 2023, https://www.forbes.com/sites/brianbushard/2023/02/24/workers-chatgpt-use-restricted-at-more-banks-including-goldman-citigroup/#:~:text=CitiGroup%2C%20Bank%20of%20America%2C%20Deutsche,has%20taken%20the%20internet%20by
6 Pranshu Virma & Will Oremus, These Lawyers Used ChatGPT to Save Time. They Got Fired and Fined., Wash. Post, Nov. 16, 2023, https://www.washingtonpost.com/technology/2023/11/16/chatgpt-lawyer-fired-ai/
7 Wolters Kluwer, Future Ready Lawyer 2023 Report, https://www.wolterskluwer.com/en/know/future-ready-lawyer-2023
8 This is the output from a prompt entered into ChatGPT.
9 Agilsium, Generative AI vs. Traditional AI: A Simple Breakdown, https://www.agilisium.com/blogs/generative-ai-vs-traditional-ai-a-simple-breakdown#:~:text=Generative%20AI%3A%20Creativity%20and%20Data%2DDriven%20Learning&text=Unlike%20Traditional%20AI%2C%20Generative%20AI,from%20vast%20amounts%20of%20data.
10 IBM, What Are AI Hallucinations?, https://www.ibm.com/topics/ai-hallucinations.
11 Xiao Fang et al., Bias of AI-Generated Content: An Examination of news Produced by Large Language Models, 14 Sci. Reports 5224 (2024), DOI: https://doi.org/10.1038/s41598-024-55686-2.
12 To learn more about how LLMs work, see Alex Freeburg & Erik Dahl, Large Language Model Fundamentals, Trial, Mar. 2024, at 46.