Aaron Crews is Chief Product and Innovation officer at UnitedLex
Undoubtedly, you’ve heard a lot of good and bad things about OpenAI ChatGPT. ChatGPT has been banned by several colleges and schools for its ability to write essays. ChatGPT’s latest version (GPT-4) passed the bar exam – and not by a small margin, but “a significant one.”
ChatGPT has been used by people for many different purposes, including writing articles*, creating emails, summarizing publications and articles, and even writing poetry, programming computer programs, and composing songs! ChatGPT was even the officiant at a recent wedding.
ChatGPT has also had significant problems. OpenAI has been sued for copyright and libel, and there have also been concerns over data privacy. The FTC announced recently that it is investigating OpenAI over potential consumer protection violations.
ChatGPT has also caused problems for lawyers trying to use it in their work. Two lawyers were recently sanctioned by a court in the Southern District of New York for using bogus case law as research to prepare a brief.
Several Federal court judges, including in Texas, Illinois, and Pennsylvania, have issued orders that require mandatory certifications from attorneys about whether they used AI, and if they did, how they verified the results. They expect lawyers using ChatGPT and other generative AI technologies to show that they have the knowledge and expertise to properly use the technology and to train and test its results.
ChatGPT Enhances eDiscovery
ChatGPT is a powerful tool that has been proven to improve workflows, increase efficiency and address common challenges for legal professionals. Legal technology companies like UnitedLex are already using this innovative method to improve their processes. ChatGPT is revolutionizing eDiscovery in many ways.
Managing Data Volume
Let’s begin at the very beginning of the EDRM Life Cycle. The explosion of digital information today can make it difficult to manage and sort through this data, especially when 85% of the data within an organization is redundant, obsolete, or trivial (ROT). ChatGPT helps manage large volumes of digital data, helping to remediate ROT and extract relevant data.
Early Case Assessment
ChatGPT is able to analyze and summarize a legal complaint. This includes identifying and listing names, dates, and other facts that are mentioned in the complaint. It can also identify and flag keywords or phrases relevant to the case. This analysis will help you make early decisions to maximize your chances of success.
ChatGPT can be used to analyze the case and compare it to other cases to gain insight into similarities and differences. Your legal team can use this type of comparison to refine strategies and strengthen arguments.
Identification of Patterns and Anomalous Anomalies
ChatGPT scans and analyses data quickly to identify patterns, trends or anomalies relevant to a particular case. It can, for example, identify unusual emails or financial transactions which could indicate fraudulent activity. This can streamline the investigative process.
In a recent webinar, I explained that with these generative text, it is essentially two algorithms (a ping pong game, if we like) playing together, where one player builds and the other validates or rejects the other’s suggestions. The process is repeated until an agreement has been reached.
Create Effective Keyword Searches
ChatGPT will help you identify relevant keywords, and construct effective Boolean search strings. The process involves engaging in a dialogue with ChatGPT for suggestions on keywords, improving search queries and creating comprehensive Boolean searches. This process of iteration can result in enhanced keyword searches which efficiently uncover key documents within the case.
Streamline Document Review
Document review is the most expensive and time-consuming phase of eDiscovery. ChatGPT helps to streamline document reviews in two ways. 1) It can generate guidelines or criteria to determine what documents are relevant to each request made in the request for producing documents. 2) It can do an initial review and identify relevant documents using criteria such as keyword searches, semantic content analyses, etc.
It’s Easier to Say than Do Without Expertise
I am Chief Product and Innovation officer at UnitedLex. As such, I know how ChatGPT supports the eDiscovery cases above. We continue to learn about its capabilities and limits to support eDiscovery. If you ask ChatGPT about how it can streamline eDiscovery it will likely identify many of the same eDiscovery cases that I’ve outlined.
ChatGPT also claims that it can assist with “comprehensive research on the law” and “finding relevant cases and legal opinions”. As we all know, the New York Sanctions case was a success.
That’s what it is all about. ChatGPT is not a “Staples Easy button” that will enhance and streamline eDiscovery. It requires training the model with machine learning techniques and then fine-tuning it to optimize ChatGPT for eDiscovery.
The platform or tool used can affect the technical implementation for passing data to ChatGPT. This is done in order to train, test and refine the system. Some common approaches are:
- Training pipelines: These pipelines usually involve scripts or commands to process and feed annotated data into the model for training.
- SDKs and APIs Some platforms and tools offer Application Programming Interfaces or Software Development Kits that allow developers to integrate ChatGPT, and perform training and fine tuning programmatically. The APIs enable data to be sent to the model in a specific format for training and fine tuning, and then the responses or updates to be retrieved and processed further.
- Custom Coding In some cases, it may be necessary to use custom scripting and coding in order to interact with ChatGPT and perform training or fine-tuning. It may be necessary to write scripts in order to preprocess data, perform API queries on the model and carry out the training and fine tuning process.
Without effective training, testing, and fine-tuning of any AI algorithm (including ChatGPT), the results will be flawed at best. ChatGPT remains a tool that needs to be properly used to produce the desired results. ChatGPT is only as good as the people who use it.
The conclusion of the article is:
Lastly, eDiscovery professionals need to be flexible in their thinking. They should have a growth mindset and think of new ways to use these tools. At UnitedLex, we can look at corporate data and use these technologies to assist clients. The tools in litigation look at your core corpus and suggest themes that you can use to convince a jury of why you should be successful. Smart practitioners will take advantage of these opportunities.
Legal professionals can achieve amazing things by using ChatGPT. It’s easier said than accomplished, but your team can achieve better results with ChatGPT and the right people, processes, and technology.
*Not this one. Except where noted, it was me!
Click here to learn more about UnitedLex Litigation & Investigation Services for Corporations.
Resources:
These schools and colleges have banned chat GPT and similar AI tools
https://www.abajournal.com/web/article/latest-version-of-chatgpt-aces-the-bar-exam-with-score-in-90th-percentile
https://people.com/chatgpt-officiates-colorado-wedding-7555337
Lawsuit filed against ChatGPT due to hallucination: Artificial intelligence trends
OpenAI is now being sued for copyright. It’s not a joke! Artificial Intelligence Trends
https://www.documentcloud.org/documents/23856993-sanctions-mata-v-avianca?responsive=1&title=1
https://www.veritas.com/news-releases/2016-03-15-veritas-global-databerg-report-finds-85-percent-of-stored-data