The Ukrainian conflict has helped expose the deficiencies of the US target recognition system known as Project Maven, veteran Russian military observer and defense tech expert Alexei Leonkov told Sputnik.
While the system in question, which is designed to use machine learning to identify targets, proved to be effective in confirming Russia’s preparations for the special military operation prior to February 24, 2022, it became overwhelmed when the fighting began and the situation on the ground started changing quickly, Leonkov explained.
“Here’s an example. An observation unit – a drone or a satellite – spots a tank. Here’s a question: is it a tank? Is it a real tank? Is it the same tank that was spotted there yesterday or is it a new tank that arrived there 15 minutes ago?” he said. “Basically, when static intelligence data suddenly became dynamic intelligence data, the Maven project faltered, which led to miscalculations during the assessment of the situation on the battlefield.”
According to him, Ukraine's “counteroffensive” last year became one of Maven’s biggest failures: the system apparently failed to take into account a “number of factors” while assessing the Russian defenses, which led to Ukrainian forces suffering a humiliating defeat where they were projected to win.
Maven also failed to anticipate the Russian attack at Avdeyevka, Leonkov said, adding that the loss of that city apparently dealt a huge blow to the system’s reputation in the Pentagon’s eyes.
Project Maven could probably be effective in “low-intensity local conflicts” where the situation on the battlefield does not change as quickly as in the Ukrainian conflict zone, he suggested.
Leonkov further noted that even US artificial intelligence experts have warned that the AI today operates within the framework of preprogrammed scenarios and algorithms, and is thus useless in the chaotic and unpredictable battlefield.
What is Project Maven
Project Maven is one of the latest and most successful efforts by the United States to integrate artificial intelligence into modern warfare.
Maven essentially uses machine learning to identify troops and military hardware by processing data obtained from various sources such as drone surveillance and satellite observation.
Ideally, this system would allow operators to quickly detect enemy troop movements and perhaps even anticipate the enemy’s actions, at a speed surpassing that of human analysts.
Currently, the United States is actively using Maven in the Ukrainian conflict, with the system being employed to provide Kiev regime forces with information on Russian troop movements in a timely manner, The New York Times has noted.
Despite the progress the US military achieved in this area, Maven is far from perfect.
For example, members of the US 18th Airborne Corps “can correctly identify a tank 84% of the time” whereas Maven “gets it closer to 60%,” Bloomberg reported last month, noting that the latter number can even drop down to 30% “on a snowy day.”
The NYT also complained that Kiev forces failed to implement one of the components of Maven – or rather, the “shadow Project Maven” analog they allegedly created – namely, the provision of “the picture of the battlefield” to the Ukrainian “soldiers in the trenches.”