The US military is exploring the potential of generative AI systems to optimize targeting decisions, with human oversight and review. This development comes amidst ongoing scrutiny of a recent strike on an Iranian school, which remains under investigation.

A Defense official disclosed that a list of potential targets could be fed into a generative AI system, which would then analyze the data and rank the targets based on factors such as aircraft location. Human operators would be responsible for evaluating the results and recommendations provided by the AI system.

Prominent AI models like OpenAI’s ChatGPT and xAI’s Grok could potentially be utilized for this purpose, as both companies have recently established agreements for their models to be used by the Pentagon in classified settings. The official emphasized that this represents a potential use case, but did not confirm or deny current usage of such AI systems.

The US military has been developing a ‘big data’ initiative called Maven since 2017, which leverages older AI technologies like computer vision to analyze vast amounts of data and imagery. For instance, Maven can algorithmically identify targets from thousands of hours of aerial drone footage.

The integration of generative AI as a conversational, chatbot layer could significantly accelerate the target search and decision-making process. However, generative AI systems are a relatively new and less battle-tested technology, built on large language models, and their application in warfare is still being explored.

Photo by Rafael Minguet Delgado on Pexels
Photos provided by Pexels