Thought Leadership

Harnessing the Power of Large Language Models: The New Frontier for Legal Tech

Published 26. June 2023.

At Semeris, we've found a playful way to refer to our analysts: AI Wranglers. Like the cowboys of old taming wild horses, we're exploring the vast landscape of AI using large models. Our goal? To discover the ideal blend of AI power and human understanding for solving real-world problems.

Unleashing the Mustangs: Large Language Models

LLMs are a distinct AI category that processes human language. Notable models like OpenAI's GPT-3 and GPT-4, Meta’s LLama, Google's PALM, and new open-source alternatives are making waves across industries (if you're curious about some AI terms, our Demystifying AI Jargon: A Semeris Guide blog post can offer some clarity).

Like wild mustangs, these LLMs are powerful and versatile but require skilled handling to harness their full potential.

Zero-shot and few-shot learning: The Initial Strides with LLMs

Large pre-trained language models hold vast knowledge, enabling novel problem-solving techniques. Zero-shot learning lets the model respond, analyze, or summarize with no guiding examples, only a simple question. One-shot and few-shot learning use a low number of examples for guidance, still producing intelligent results.

Despite their prowess, these models can hallucinate, producing outputs unrelated to their training or missing the mark on questions or examples. Much like a startled horse, they can create inaccuracies that need the intervention of an expert.

AI Wranglers: The Good Sheriffs of AI

In the vast AI landscape, the role of an AI wrangler is critical. Our experts tame LLMs, meticulously crafting queries, and fine-tuning models for accuracy. They create ground truth datasets for evaluating outcomes and are well-versed in prompt engineering - the emerging art of interacting with LLMs.

Their responsibilities, however, extend beyond precision and accuracy. They ensure that our AI respects privacy, maintains fairness, and avoids bias, a crucial aspect of our operations.

Similar to a vigilant sheriff maintaining the safety of a frontier town, our wranglers also prioritize the security of our AI operations. Our commitment to data protection is demonstrated by achieving SOC 2 Type II Compliance.

By upholding ethical standards and ensuring robust security, our AI Wranglers deliver not just powerful AI, but AI you can trust.

A Hybrid Approach: Combining AI with Human Insight

At Semeris, our AI wranglers (assisted by Semeris X-Ray) use a hybrid quality assurance approach, blending secondary AI reflection (where a second AI checks the outcomes of another AI) with human oversight, similar to a cowboy guiding his horse by understanding its capabilities.

With customers managing trillions in assets, there's no room for unrestrained AI, just as a cowboy can't risk an unbroken horse. Our experts are vital, refining AI's raw power to ensure optimal precision.

Riding Towards an Efficient Frontier with Semeris

Like wild horses, LLMs are powerful and beautiful but demand expert handling. At Semeris, we're prepared to guide you through the intricate AI terrain. We invite you to reach out if you face questions or immediate analytical issues. We're keen to customize our expertise to suit your needs.

Connect with us today, and let's tap into AI's power together.

Welcome to the efficient frontier!