Download

Download Full Text (7.8 MB)

Description

Intelligent Tools for Designing Operational Concepts: Intelligent Agents Innovating Mechanisms for Victory

If a large model is likened to a machine, interacting via conversation is akin to a worker manually operating that machine. Although machine tools have significantly reduced human workloads, “manual operation” still entails challenges: demanding skill requirements for operators; difficulty in resolving complex, specialized tasks through conversation alone; difficulty in standardizing and reusing repetitive workflows; susceptibility to hallucinations; and the time- and labor-intensive nature of verifying results. To this end, the Department of Theoretical Innovation Research has proposed a technical roadmap for programming large models. The underlying concept of this technical approach is to use programs to drive large-scale models in executing complex, specialized tasks, effectively upgrading them from manual machine tools to computer numerical control (CNC) machine tools and, furthermore, integrating multiple CNC machine tools into a “production line” capable of complex workflows. This significantly enhances the capability, efficiency, and speed with which large models can address complex, specialized issues in the defense and military sectors, while also preventing or eliminating low-quality responses and “hallucinations” that fail to meet mission requirements. The result is a more user-friendly system that achieves true “domain-specific intelligence.”

Publication Date

8-31-2026

Publisher

China Maritime Studies Institute, U.S Naval War College

City

Newport, Rhode Island

Keywords

China Maritime Studies Institute, CMSI, PLA, PLAN, PLA Navy, AI. LLM, Large Language Model, Machine tools, Machine learning, hallucinations, butterfly effect

CMSI Translations #35: How to Leverage the

Share

COinS
 

Accessibility Request

Some items in this repository were created or digitized prior to implementation of the accessibility standards under the Rehabilitation Act of 1973 and are preserved in their original, unmodified state for research, reference, or historical recordkeeping. In accordance with the ADA Title II Final Rule, the College provides accessible versions of archival materials upon request. To request a version of a file or resource, please submit an Accessible File Request Form.