Background
Figure1: LAMP logo.
This project supports a Focused Internal Research and Development (IR&D) Program to explore and harness the capabilities of Large Language Models (LLMs), which are artificial intelligence designed to understand and generate human-like content, including text and images. These models, built on extensive datasets, are already leveraged for tasks such as translation, content creation, and query response by interpreting context to generate relevant outputs. This initiative focuses on advancing technological capabilities and uncovering new applications that can address real-world challenges and contribute to societal progress.
Approach
The program had an initial call for proposals across the Institute and selected ten projects across technical domains such as transportation, defense solutions, biomechanics, chemistry, robotics, and probabilistic analysis of structural components. Demos and presentations to SwRI staff highlighted the lessons learned and the technical expertise gained. Initial projects may submit proposals for follow on projects and a second round of investigative projects will be awarded at the end of this year. This approach emphasizes collaboration and knowledge dissemination among funded projects and interested staff, thereby increasing efficiency and highlighting LLM's potential impact on operations. This program promotes cross-Institute communication and fosters innovation in LLM development and application and seeks to establish guidelines and best practices for safe and effective LLM accessibility for SwRI employees.
Accomplishments
Projects used a range of LLMs from the closed-source GPT-4 to open-source models such as Llama and Mistral, with many projects incorporating Retrieval Augmented Generation for a variety of tasks. Automation was a core output of most of the projects, although the level of automation varied from project to project. An interesting topic illustrated by two of the projects involved workforce resiliency. With the retirement of subject matter experts, companies are at risk of losing domain knowledge critical to their business. Projects explored using LLMs to preserve knowledge and maintain continuity of operations. SwRI will continue to explore the ethical use of LLMs to further our mission to benefit government, industry, and mankind.