LLMs & The Future of the Legal Profession

This is the introductory lesson, where we quickly go over the reasons why this course is topical, and why LLMs are likely to become so important for lawyers going forward.

Explains how text is transformed into data in modern NLP pipelines, introducing concepts like word embeddings and showing how these foundational methods enable machines to parse, compare, and generate human language at scale. In short, how we made text commensurable, unlocking its potential.

Describes the move from older neural network approaches to the Transformer architecture, highlighting the importance of attention mechanisms, tokenization, and autoregressive text generation for state-of-the-art language models.

Shows how LLMs can be customized for specific tasks through fine-tuning and prompt engineering, discusses RLHF (reinforcement learning from human feedback), and explores how advanced “agent” architectures orchestrate multiple AI capabilities.

Details the principal shortcomings of LLMs - including bias, hallucinations, and lack of grounding - while also exploring their transformative possibilities when paired with retrieval techniques and human oversight.

Reviews how scaling laws, data availability, and computational resources drive LLM performance, addressing related challenges such as high costs, energy consumption, and geopolitical concerns over AI development.

Surveys how LLMs will reshape legal tasks, workflows, and career paths, showing where human expertise remains vital and how lawyers can partner with AI to drive new forms of value creation.

Investigates the impact of AI on creative tasks and legal practice, introducing the concept of the “centaur lawyer” who wields LLMs for efficiency while retaining human judgment and expertise in shaping legal strategy.

Explores how LLMs might transform legal research, decision-making, and access to justice, spotlighting issues of fairness, bias, and transparency in an AI-driven legal system.

Will we all die because of AI ?