This article was originally published on October 25, 2023 on Legal Tech Hub. They are the original source of this article and it can be found here. 

In the realm of legal technology, Richard Robbins stands out for his extensive experience in both law and AI. Currently, as the Managing Director of Applied AI at Epiq Global Legal Solutions, Richard plays a crucial role in incorporating AI into legal practices. Legaltech Hub interviewed him to gain insights into the application of Large Language Models (LLMs) in law, discussing his background, Epiq's initiatives, and how law firms are and should be utilizing AI.

Richard’s Background 

Richard's journey began over 45 years ago in computer science. He studied at MIT, where he researched topics related to AI, and subsequently worked in the space for several years. After a career pivot that led him to earn his Juris Doctor degree from the University of Chicago Law School, Richard practiced law for 24 years, 14 of which were spent at Sidley Austin, ultimately as a partner. He also worked in-house, serving as the General Counsel at Morningstar for a decade. After leaving Morningstar, Richard returned to Sidley to establish the firm’s knowledge management department. Post-Sidley, Richard joined his colleagues at Fireman, which eventually became a part of Epiq.

In the midst of his career, Richard decided to further his education in data science, enrolling in a Master's program at the University of California, Berkeley, with a focus on machine learning and natural language processing. Notably, during this period, technologies like ChatGPT were emerging, and Richard became part of the educational team specializing in natural language processing and ChatGPT-related technologies.

Given his background and his recent engagement with Large Language Model (LLM) projects, including text generation, it was a natural move for Richard to transition into Epiq's corporate parent as the Managing Director of Applied AI.

What services are you building out with the Epiq AI group? 

The Epiq AI group is not a standalone product group but rather a consortium of experts from throughout the Epiq organization, each possessing relevant AI expertise and experience. This unique practice group integrates technical specialists whose work intersects with various sections of the organization, thereby influencing and enhancing the services provided.

The group’s work influences the development of internal platforms used for delivering solutions, aiding corporate legal departments and law firms. Currently, the group is contemplating a wide array of use cases for generative AI and working on multiple projects aimed at harnessing evolving AI technologies to best serve their clients.

Whether a client approaches the Applied AI Group with a specific need like eDiscovery or contract review, the group aims to employ the most advanced and effective technology to yield optimal results. This approach could involve utilizing third-party technology or enhancing Epiq's proprietary technology, depending on the project's demands. Some of these solutions are offered directly to clients through the Epiq Service Cloud or various advisory services.

How are law firms using LLMs?

LLMs are gaining traction in legal firms, mainly for their capability to interface with internal systems and generate new content.

Some pioneering firms have already made public their adoption of next-generation AI-based platforms, primarily as a strategy to keep sensitive data within the firm’s confines while providing their teams with access to innovative tools like ChatGPT. A trend among these early adopters is the utilization of LLMs not merely as generators of new content but also as intuitive, language-based interfaces that facilitate access to a variety of internal systems within law firms. These firms are capitalizing on the capabilities of generative AI to establish connections between disparate systems and provide seamless access to various resources. The applications of LLMs for law firms span both the business and practice sides of law firms, indicating a wide range of potential use cases.

Richard emphasizes a strategic approach to LLMs.. It's essential, he says, to understand LLMs' limitations and set realistic expectations, combining their innovative capabilities with traditional problem-solving methods.

What should law firms and corporate legal departments be doing with LLMs?

Richard emphasizes a strategic approach to LLMs. Law firms and corporate legal departments focus on solving specific business problems that are vital to their organizations instead of pursuing technology for its own sake. This requires thoughtful deliberation on what approach best suits their organization's needs and circumstances, whether that entails building proprietary solutions or purchasing available products in the market.

The technology should serve as a complementary resource, empowering professionals to accomplish their tasks with greater speed and precision rather than acting as a substitute for human expertise. The power of LLM technology is best harnessed when it is utilized by skilled individuals who can leverage its capabilities to deliver superior results.

Setting realistic expectations regarding the strengths and limitations of LLMs is also essential, Richard says. Users should not anticipate that the technology will intuitively understand and execute vague or overly broad directives. The optimal use of LLM technology combines the novel capabilities of these models with tried-and-tested methods, providing a balanced toolkit that includes both innovative and 'classic' approaches to problem-solving.

Is building the right way to go?

Most law firms, Richard points out, aren't equipped to develop software. While the emergence of low code and no code platforms has helped to democratize the creation of software solutions, law firms should focus on leveraging existing technologies rather than attempting to develop their own.

Instead of embarking on the intricate and demanding process of developing their own AI solutions, law firms would be better served by acquiring and customizing existing large language models to suit their specific needs. Engaging with consultants, advisors, and internal experts can guide law firms in seamlessly integrating these technologies into their operations.

Fine-tuning an LLM with a layer of law firm data   

The discussion around whether law firms can gain a competitive advantage by fine-tuning Large Language Models (LLMs) with their own data hinges significantly on what is meant by "LLM". For instance, while chat GPT and similar technologies are built on LLMs, there were various other large language models preceding them designed for fine-tuning. Traditional or "classic" LLMs, which are not as expansive as the more recent generative models, often serve as a base solution proficient in a few tasks. Firms could potentially fine-tune these models by layering on their own data, thus creating a tool purpose-built for specific functions.

However, the advent of generative models has changed the dynamic. These newer, expansive models are quite efficient straight out of the box, albeit not as specialized as purpose-built tools. The distinction in performance between a generatively fine-tuned model and a specialized, purpose-built tool might be marginal, thereby raising important considerations for law firms. Is the effort and resources required to fine-tune a generative model justified by the marginal improvement in performance it offers over out-of-the-box solutions?

For example, in some studies, Richard explained that while GPT-3.5 performed admirably, it did not quite match up to the performance of purpose-built models. However, these purpose-built models necessitated dedicated development, unlike GPT-3.5, which is ready for use off the shelf. Hence, law firms and organizations must weigh whether the incremental improvement offered by custom-built solutions is worth the considerable effort and data required to develop them. Essentially, they need to evaluate when "good enough" truly is sufficient to meet their needs, taking into consideration the specific demands and strategic goals of their operation.

The “wait and see” approach

Richard believes that the viability of adopting a "wait and see" approach towards the implementation of technology like GPT-4 largely depends on a law firm's positioning within the legal landscape and its specific practice areas. For clients of law firms, the emphasis is predominantly on the results the firms deliver rather than the technology they employ. Especially in cases where the work isn't commodified, clients tend to prioritize results over the intricacies of the technological tools used in the service delivery process.

However, for practices that are significantly efficiency-driven and handle high volumes of commodity work, the urgency to incorporate new technologies may be more pressing. In such contexts, these technologies are not just tools but pivotal elements that can substantially influence efficiency and output.

Richard also emphasizes that the risk for law firms isn't exclusively tied to the timely adoption of cutting-edge technologies but is also deeply connected to their data strategies. Firms that lag in organizing their data or lack a coherent, preferably cloud-centric, data strategy may find themselves at a disadvantage. The importance of acting promptly in this aspect cannot be overstated, as unlike certain commodities that improve with time, the challenges tied to data management and strategy won't simply resolve or diminish as time progresses. In this regard, procrastination exacerbates the problem.

What impact will advanced AI have on the legal market? 

The advent of sophisticated AI technologies has initiated a wave of accelerated change within the legal sector, compressing cycles that would conventionally take years into mere months. However, Richard notes that not all tech buzz translates into immediate, substantial impact on the market. For instance, while there was considerable chatter about blockchain, its fundamental influence on the legal market has been relatively limited so far. In contrast, AI's presence and integration into various solutions, especially in legal research, are undeniable and growing. These AI-enhanced solutions are not only available but actively being utilized in the present day.

Incumbent service providers in the legal industry who choose to disregard the ongoing technological shifts risk obsolescence. Ignoring the tidal wave of AI could be perilous for those resting on past laurels.

Conversely, new entrants, lured by the potential of the AI gold rush, also face risks if they misinterpret the market dynamics. Without a nuanced understanding of the market, these newcomers might find themselves ill-prepared for the industry's complexities and demands.

Therefore, the unfolding AI epoch in the legal sector demands careful and thoughtful consideration from all stakeholders. Each group must navigate through this transformative period judiciously, weighing their options and strategizing to harness the power of AI effectively and sustainably.