The US military is testing generative AI’s potential in operational efficiency, marking a significant shift in defense technology.

The US military’s latest venture into the world of artificial intelligence marks a significant turning point in its technological capabilities. Col. Matthew Strohmeyer of the US Air Force recently led an experiment that successfully integrated large-language models (LLMs) into military operations. This experiment is part of a broader initiative by the Pentagon’s digital and AI office to enhance data integration across military divisions.

Traditionally, data requests in the military involve cumbersome, time-consuming processes. LLMs promise a transformative solution. In one trial, an AI tool completed a request that would usually take hours or days in just 10 minutes. While these systems are not yet fully operational, their potential is clear.

The military is testing various LLMs, including Donovan, a product from Scale AI, a San Francisco-based startup. The objective is to incorporate AI in decision-making processes, sensor operations, and potentially in firepower. This aligns with the Pentagon’s long-term vision to modernize its operations through AI integration.

Several companies, such as Palantir Technologies and Anduril Industries, are developing AI decision platforms for the Pentagon. Furthermore, Microsoft Corp. has made AI models from OpenAI available through its Azure Government cloud service, which the Defense Department uses.

These tests also include plans to utilize LLMs in strategizing military responses to escalating global crises. However, there are concerns about AI’s reliability, such as potential biases and misinformation. Recognizing these challenges, the Pentagon is collaborating with tech security firms to evaluate the trustworthiness of AI systems.

A recent demonstration involving 60,000 pages of open-source data, including US and Chinese military documents, highlighted both the capabilities and limitations of these AI tools. The exercise underscored the need for direct US intervention in a hypothetical conflict scenario, while also acknowledging the uncertainties surrounding such military engagements.