Project Maven

Main article: UAVs in the United States

Project Maven is one of the largest American projects in the field of creating military drones with AI. This is the Pentagon's program to build an AI-based surveillance platform for unmanned aerial vehicles. The challenge is to create a system for the US military to deploy and monitor autonomous drones.

The purpose of this system is to control combat operations in real time and track, highlight and open targets without the participation of the operator. According to information from open sources, it can be assumed that at the beginning of 2024 the project is not being developed as a completely autonomous weapons system with AI, capable of independently selecting and hitting targets during hostilities. But in the future this cannot be done.

2024: Testing technology in combat operations in Ukraine for attacks on refineries in Russia using Starlink satellites

In 2024, full-scale testing of AI technologies in a real conflict in Ukraine is underway.

Project Maven processes huge amounts of data from countless sources - from satellites to social media feeds of Russian military personnel. The volume of information tens of thousands of times increases the ability of people to analyze it, which requires the use of AI.

In an interview with the NYT, Lt. Gen. Christopher T. Donahue, commander of XVIII Airborne Corps, stated, "It became our laboratory after all."

One of the main goals of Project Maven is the development of an automatic target detection system, which can then be combined with drone technology to create swarms of autonomous kamikaze drones capable of detecting and destroying targets with minimal human operator participation.

When analyzing different types of images, the operator can correctly identify the tank in 84% of cases. Project Maven AI successfully performs the task at almost 60%, and in snowfall conditions this figure drops to 30%.

Attack drones that were launched from the territory of Ukraine at Russian refineries were aimed at the target using a fairly simple version of AI based on. machine vision A source CNN said that on April 2, 2024, some of these drones began to be used, machine vision which helps drones accurately aim at the target and makes aircraft noise-resistant.

According to the source, accuracy under conditions of interference is provided by processing images from the camera and matching with data in memory. Each aircraft is equipped with a processor into which data from the satellite and the terrain map are pre-loaded. The devices follow a predetermined route, which allows you to hit targets with an accuracy of a meter. This means that the drone does not need contact with satellites.

Technology is constantly being improved and relies on US intelligence. Combat feedback allows American specialists to get information about the weak points of AI.

While the Pentagon claims that Project Maven is only a shoot (in 2017, congressmen were voiced this particular version, for funding), any developer will tell you that drone training to shoot at anything works roughly the same, whether it carries a camera or a weapon, "Fisherman" wrote.

The most important conclusion in the field of drone use appeared on the pages of the NYT: "Officials Pentagon now understand, as never before, that the American system of military satellites should be built and configured in a completely different way, with configurations that are more like constellations of small satellites." Starlink Elona Musk

Previously, the concept of using drones relied on the use of communication channels using repeaters and data from high-orbit satellites. The conflict in Ukraine showed its inefficiency. Now the Americans are focusing on the development of a much larger and more budgetary constellation of low-orbit satellites, based on Elon Musk's Starlink.

The fundamental difference between the use of low-orbit satellites is in the higher resolution of shooting and more accurate analysis of data from sensors in the IR range, technology lidars and using other remote sensing systems. Lands

If you greatly simplify, then work is underway to create a spherical analogue of the solid field location station, where all objects inside the sphere are constantly in the detection zone. A large number of low-orbit satellites (talking about tens of thousands) make it possible to ensure a constant and high coverage density. The task of the satellites is a detailed surface scan to create route tasks (attack corridors) for UAVs.

This should allow loading the current flight mission into the drone memory, taking into account changes in the terrain and/or location of the targets. For example, the air defense calculation changed position within 4 hours before the flight of conventional drones. For high-orbit satellites, this change in position went unnoticed, but the array of low-orbit devices monitors the situation in real time, which allows you to quickly update the geolocation data.

This is exactly what the US military considers a technology that will gain a battlefield advantage in upcoming conflicts.

For the United States, this is a critical technology of the future, before which countries with outdated air defense systems or without them will be defenseless. This is a pressure lever comparable to nuclear weapons, with minimal collateral damage.

As countermeasures, the logical step is to develop and launch its own grouping of low-orbital satellites and to force the development of anti-satellite weapons, in parallel with the improvement of hypersonic systems. The enemy's low-orbit grouping may be integrated into a system for intercepting hypersonic missiles in the future.

2019: Google withdraws from the project. Palantir gets development contract

Founded by Peter Thiel, Palantir has filled the void created after Google dropped Project Maven for ethical reasons in early 2019. In the media, a whole performance was played on this topic.

Peter Thiel, the founder of Palantir and a longtime associate, Donald Trump called Google's withdrawal from the project tantamount to betrayal. In June 2018, speaking at the National Conservatism Conference in, Washington Thiel mentioned Google's decision to promote Project Dragonfly (Google's program to create censored search for), China while abandoning Pentagon the Project Maven project. He said he CIA should investigate Google.

Inside Palantir, where the privacy of customers who have gained access to the company's technology is tightly controlled, the project is called "Tron," after Steven Lisberger's 1982 film.

Tech companies must do their patriotic duty, according to Thiel and Palantir CEO Alex Karp. And for this it is necessary to do absolutely everything that the US government will say.

2017: Start of the project with the participation of Google

The Maven project started in 2017 and is the main attempt Pentagon to introduce promising AI developments for war so that algorithms make decisions instead of humans.

Initially, Google received this large-scale contract. In 2018, about a dozen employees staged a small vigil after Google Cloud CEO Diane Green announced the company would renew the contract. This gave a formal reason to Google management in the future to refuse to re-participate in the tender.