Data is to generative AI what roads are to cars

A review of Karl Chapman’s session at Legal Geek 2023

Karl Chapman, CEO at Kim, recently took to the stage at Legal Geek in London to share his insights into generative AI. During his thought-provoking session using interesting analogies and dubious images, Karl underscored the importance of discernment in an environment inundated with misleading information and what he aptly termed 'hallucinations' about generative AI.

Drawing a parallel between this field and the notoriously unpredictable nature of forecasting, Karl delivered a stark message: distinguishing between the known and the unknown is pivotal when wading through the complexities of generative AI.

Karl's session wasn't just about theoretical discourse. He dove deep into practical aspects, focusing on demonstrating the value of one's role, the effective management of risks, and harnessing the potential of generative AI. At the heart of it all, he unveiled the session's central question: What's the key to success in this fascinating domain? The unequivocal answer, he asserted, was data. As he humorously pointed out, the hint was right in the session's title.

The Role of Data in Navigating the Generative AI Landscape

As the session progressed, the spotlight shifted to the vital role of data in the generative AI landscape, emphasizing the transformation of raw data into actionable insights. In particular, the legal context emerged as a prime example of where data serves as a trusted ally, providing guidance and protection. Karl shed light on the crucial role of intake management in legal operations, emphasizing that clear, accessible data streamlines processes, ensures accurate instructions, and facilitates the tracking of case progress.

The generative AI journey was a key subject of discussion. Karl even indulged in a touch of humour, alluding to a historic football event [World Cup 1966 and the fact that generative AI proves the ball crossed the line] to underscore the undeniable impact of generative AI. The audience's experience with chat GPT and related plugins was explored, setting the stage for the transformative potential of generative AI, particularly in tasks like document drafting and content creation. Nevertheless, Karl cautioned that careful application selection was essential in harnessing its capabilities effectively.

The Data-Driven Future lies in Private Legal Language Models (LLMs)

The session then shifted its focus to a vision of tomorrow's successful organizations, marked by their possession of private and secure Legal Language Models (LLMs). These LLMs are not exclusive to the legal function; they are set to be organization-wide. Central to this vision is the quality and accuracy of the data fed into these models, and Karl's message was unequivocal: "garbage in, garbage out." The critical importance of pre-processing data for accurate and deterministic outcomes was highlighted, with an example illustrating the need for consistent responses to requests, such as those for a German distribution agreement.

Karl employed the metaphor of the gold rush to drive home the idea that data is the true treasure, the lifeblood of LLMs. He stressed the necessity of re-engineering existing data to optimize LLM performance.

The Plugin Solution

In the context of future prospects, Karl discussed the integration of plugins as a solution for organizations. This integration would ensure uniform, organization-specific responses to prompts, preventing multiple variations of the same request and providing a unified approach. While acknowledging that tactical problems can be solved today, he emphasized that the strategic focus must remain on data and its capture methods.

As the session ended, Karl's parting message was a call to embrace change. Data, he emphasized, was foundational for any organization, and he warned against attempting to fit generative AI into outdated data models, as it would likely result in inefficiency. The session concluded enthusiastically, with Karl reiterating a simple but powerful sentiment: "We all love data!" In his words, data is to generative AI what roads are to cars - an essential and invaluable resource that propels us into the future.