Artificial intelligence aboard an American spy plane took control of the flight

WASHINGTON, (BM) – We often ask ourselves whether the films categorizing as “science fiction” are such and not reality. There are already many examples that would disprove the genre of these films and put them in another. However, not as a fantasy film, but the reality was the flight of the U-2 spy plane of the US Air Force on December 15. Back then, the co-pilot was an artificial intelligence system that took control of the spy plane’s radars and sensors, learned citing Valerie Insinna for Defense News.

Read more: The United States will fight Russian armored vehicles with artificial intelligence

The co-pilot is the Artuμ artificial intelligence system. The flight on December 15 was an exercise to detect and repel an attack by “enemy” missile launchers. Artuμ’s goal was to locate them quickly and efficiently to help the first pilot destroy them. Artuμ also controlled the aircraft’s tactical navigation after it took off from the runway. The whole exercise was planned. It took place at the Biel Air Force Base in California.

Air Force Acquisition CEO Will Roper says of the flight that artificial intelligence is a necessity. According to him, if the US military wants not to be history, it must become “science fiction.” Roper also says that the artificial intelligence system made excellent decisions during the exercise, but there are strengths and weaknesses like all of them. Roper did not want to specify the shortcomings of the “co-pilot.”

Roper provided more information to the media about the Artuµ artificial intelligence system. According to him, the software engineers created the AI system in the laboratory conditions of the U-2 program. The source is a game chess algorithm, which had previously used in a human-machine competition. In October this year, the U-2 updated the software using an open-source containerized method. When the procedure ended with a positive result, the engineers of the U-2 program set the manipulative actions of the “second pilot Artuµ” to influence the sensor package of the aircraft. According to Roper, before the test flight and training, Artuµ played more than half a million computer simulation missions.

According to him, Roper added that control over the radar and sensor systems of the U-2 spy plane is only part of Artuµ’s task. In the future, this artificial aircraft will take over and control other functions during the mission. Artuµ also can be integrated into other US Air Force fighter jets. Roper concludes by saying that digital superiority is the key to success in future wars, and Artuµ gives a significant advantage to the US military.

“AI will play a critical role in achieving that edge, so I’m incredibly proud of what the team accomplished. We must accelerate change, and that only happens when our Airmen push the limits of what we thought was possible,” says Roper.

Four US companies will develop artificial intelligence for the military

The four US companies Boeing, EpiSci, Heron Systems, physicsAI and the Georgia Institute of Technology were selected by the Defense Advanced Research Projects Agency [DARPA] to develop artificial intelligence for the US military, as we reported in November.

According to information from the site Popular Mechanics, the contract has already been signed and the selected companies and technical institute will have to develop a system of artificial intelligence that will allow in the near future to support American aviation by conducting independent air combat and close combat.

Read more: The US believes artificial intelligence will give the Pentagon a military advantage

The idea of ​​artificial intelligence in military affairs is not recent, but is embedded as a concept in the most militarily developed countries. We are witnessing the United States, Russia and China working hard in this area, and many other countries have military weapons systems that are at the penultimate stage to artificial intelligence – autonomous weapons systems.

According to the set characteristics of DARPA, the future artificial intelligence system must be self-taught and draw logical conclusions from its mistakes. According to Pentagon sources, the artificial intelligence system should initially enter the US military as an integrated system in certain flight systems, which will be a supporting force for real pilots during missions. Subsequently, the system must allow the aircraft or a particular weapon system to be able to conduct three types of direct combat – one against one, one against two and two against two.

The military logic leads to the conclusion that after developing a system with artificial intelligence, the military will take a long process before integrating the system, and the system itself will go through at least two stages of testing – work in a simulated environment and work in a real environment. The concept is that an AI weapon system would have a shorter response time in a particular combat environment, as well as suggest more options for solving a problem or winning a particular battle.

We remind you that only months ago in the United States a similar battle took place between a real pilot of the US Air Force against a neural network system, and the final stage of the battle was quite disappointing for the man, as he lost all three stages of the race with a score of 5: 0. each stage for the machine. Perhaps for this and other reasons, the artificial intelligence system will initially be targeted for integration into the US Air Force, with mini-and medium-sized drones expected to be the first weapon systems to acquire it.

By the end of 2023, the United States is expected to have an artificial intelligence system that has passed all possible tests and is ready for integration.

Currently, most weapon systems in the world are autonomous. This means that any autonomous weapon system can connect to the nearest friendly system and help win a battle. There are many such examples, the most noticeable in the last few months is the combined flight between the Russian Su-57 fighter and the Hunter drone, which is actually connected to the systems of the Su-57 pilot and appears as the so-called “extended wing” of the fighter.

However, in the presence of such a drone with artificial intelligence, not only will the drone be in full symbiotic connection with the communication systems of the fighter and the pilot, but will also be able to decide for itself how to provide real support to the pilot, or faster than him to be able to locate a desired target and destroy it, protecting the life of the pilot and expensive aircraft.

Artificial intelligence is already entering the weapons systems of land-based military units. In Israel, the military of the Jewish state has developed an entirely new tank with artificial intelligence, which is borrowed from the artificial intelligence of the electronic game Warcraft 2. The project is so successful that it is currently used to train recruits who have recently been part of the civilian life. It turns out that the training time for tank recruits is drastically reduced and significantly – from several months to several hours.

Read more: US will use artificial intelligence of the ABMS systems to help fight Russian missiles


Follow us everywhere and at any time. has responsive design and you can open the page from any computer, mobile devices or web browsers. For more up-to-date news from us, follow our YouTube, Reddit, LinkedIn, Twitter and Facebook pages. Do not miss the chance to subscribe to our newsletter. Subscribe and read our stories in News360App in AppStore or GooglePlay or in FeedlyApp in AppStore or GooglePlay

Subscribe to Google News

>>Be a reporter: Write and send your article.<<
Editorial team