If you’ve been keeping tabs on the advancements in the realm of Generative AI, you’d know that the terrain has been in a constant state of upheaval since the public release of ChatGPT 3.5 on November 30, 2022. Navigating this exhilarating but often overwhelming landscape could be a bit of a challenge for legal professionals. It’s a dynamic space, teeming with competition, new entrants, and continuously evolving trends.
Once dominated by ChatGPT, the market is now seeing a surge in Large Language Models (LLMs). Companies like Google, Meta, and Databricks have each introduced their respective LLMs BARD, BERT, Llama 2.0, Falcon, and Dolley. Particularly noticeable is the emergent focus on domain-specific LLMs that can cater effectively to sectors such as engineering, medicine, or law. Start-ups like Harvey.AI and Alexi.AI exemplify this trend, targeting legal applications with Generative AI.
Revisiting the early days of computing when companies resorted to time-sharing due to economic constraints illuminates the current stage of Generative AI. Likened to the time-sharing era, the present phase is witnessing a monopoly by major players like OPENAI/Microsoft, Google, and Meta. However, there’s anticipation of progression into the “personal computer” era soon where companies and law firms could build their own proprietary Generative AI systems.
The competitive environment and fast-paced developments are expected to make Generative AI more economical. Factors like the availability of AI-specific Graphics Processing Unit (GPU) technology play a part in driving costs down. Not every application will require 175 billion parameters like ChatGPT 3.5 Turbo, opening the door for smaller models designed for specific data sets and problems. In fact, a report by Goldman Sachs earlier this year even endorsed significant productivity enhancements resulting from substantial investment in AI.
Another promising development in Generative AI is the Retrieval-Augmented Generation (RAG). Particularly useful for professional organizations, like law firms, that can’t afford to err, RAG ensures a minimal margin of error. Tools like ChatGD by Gunderson use this technique. Fundamentally, these solutions leverage natural language search to find relevant documents in a database before using Generative AI, like ChatGPT, to summarize these.
Having largely explored the unfolding landscape of Generative AI, it is evident that although there will be challenges, legal professionals who embrace and understand this technology will likely have the upper hand. For an in-depth exploration of these topics, you might find ‘How Generative AI Works (Part IV)‘ by Ken Crutchfield, Vice President and General Manager of Legal Markets at Wolters Kluwer Legal & Regulatory U.S., a compelling read.