Separate Registration Required

Large Language Models (LLMs) such as ChatGPT, Claude, and Gemini have revolutionized how information is processed and generated. However, aligning these models with the specialized reasoning and decision-making processes of GEOINT analysts remains a critical challenge.

This session explores methods for adapting and refining AI models to mirror expert-level decision attributes within the GEOINT domain. Drawing from the DARPA In the Moment (ITM) program, participants will see real-world demonstrations of model alignment in high-stakes contexts such as battlefield medical triage and cybersecurity defense, illustrating how similar approaches can be applied to GEOINT analysis.

Attendees will also learn how to fine-tune and prompt-train recent LLMs using domain-specific data, evaluate performance across multiple models, and explore extensions into multimodal AI, including imagery and speech generation (e.g., Stable Diffusion for satellite imagery).

Learning Outcomes

Understand the strengths and limitations of major LLMs (ChatGPT, Claude, Gemini) for GEOINT use cases
Explore methods for adapting and aligning AI models to expert decision-making processes
Gain familiarity with tools for local and domain-specific model refinement (e.g., Ollama)
Learn how generative models across text, imagery, and speech can enhance GEOINT analysis

Prerequisites

No formal prerequisites.
Prior exposure to LLMs such as ChatGPT, Claude, or Gemini will enhance learning.
Familiarity with tools like Ollama or local model environments is helpful but not required.

GEOINT 2026

Alignment of AI with Decision-Making Attributes of GEOINT Experts

Type

Speakers

Description

Tracks