AI Accelerates War: Targeting in Seconds

The speed of modern warfare has been significantly accelerated by artificial intelligence, as detailed in a new book. The US military's assault on Iran, for instance, saw over 1,000 targets struck in its first 24 hours, nearly double the scale of the 2003 Iraq invasion.

Central to this acceleration is the Maven Smart System, which synthesizes satellite imagery, radar, social media, and other data to rapidly identify targets. Originally conceived in 2017 as an experiment using AI on drone footage, the project faced initial controversy, including employee protests at Google, the original contractor.

Despite early setbacks, the system, driven by Marine intelligence officer Drew Cukor, was developed with contributions from Palantir, Microsoft, Amazon, and Anthropic. Maven combines computer vision with a workflow system, reducing the time to target from hours to seconds. Officials report the capability to strike up to five thousand targets daily with LLM integration.

The rapid pace has led to tragic outcomes, such as the erroneous targeting of a girls' school. While chatbots like Claude faced scrutiny, experts emphasize that the accelerated targeting process enabled by systems like Maven is the more critical factor behind lethal failures.

Military programs are further advancing towards fully autonomous weapons, indicating an ongoing, rapid evolution in AI warfare. This includes autonomous drones capable of independent targeting and destruction.

The development of Maven has seen significant shifts, with Palantir eventually becoming the prime contractor for the Maven Smart System, which is set to become a program of record. Ukraine's conflict served as a critical inflection point, showcasing how AI could speed up artillery operations and targeting by analyzing Russian positions in near real-time.

This AI integration has reduced human involvement in the targeting cycle to key decisions like the initiation and execution of strikes. Even intelligence reports are now being generated entirely by AI, underscoring a profound shift towards data-centric military operations.

Concerns linger regarding the potential gamification of war and the risk of over-reliance on AI-provided targets without fully understanding the supporting data. However, proponents argue that AI systems offer greater transparency and accountability by making data auditable.

The military is actively debating the extent to which it should embrace AI in targeting, with some viewing it as inevitable and others emphasizing the continued importance of human assessment to prevent errors and save lives. Historical incidents, like the 1999 bombing of the Chinese Embassy in Belgrade due to an outdated map, highlight the risks of faulty data, regardless of system speed.