Home » knowledge » Chinese pilots battle AI opponents in simulated dogfights

Chinese pilots battle AI opponents in simulated dogfights

Chinese pilots battle AI opponents in simulated dogfights

A recent report by Chinese state media said that in simulated dogfights, Chinese Air Force pilots lost considerable time in front of AI-powered adversaries. This sounds reminiscent of the very public results of last year’s DARPA’s AlphaDogfight trial, which has since been used for more advanced demonstrations. It also highlights the PLA’s growing interest and investment in the development of advanced AI and machine learning technologies in general.

Earlier this week, Chinese state media Global Times reported on the People’s Liberation Army Air Force (PLAAF) fighting “artificial intelligence aircraft” in a simulator, citing another report in the PLA Daily over the weekend.

Chinese pilots battle AI opponents in simulated dogfights

Xinhua News Agency: The Chinese People’s Liberation Army Air Force pilots were assigned to the Bayi Aerobatic Team.

“The AI ​​demonstrated adept flight control skills and accurate tactical decision-making, making it a valuable adversary for honing our capabilities,” identified the commander of an unspecified PLA Air Force brigade assigned to the PLA’s Central Theater Air Force. Du Jianfeng told PLA Daily, according to the Global Times.

The Global Times report also said that AI has reportedly been used in simulator training for “years” and that it is able to “learn from pilots because it collects data from each training session”. Therefore, “At first, it was not difficult to defeat AI. But by studying the data, every engagement became an opportunity for it to improve,” Fang Guoyu, the brigade leader of Du’s unit, also told the PLA Daily. In a recent real-world air combat exercise, Fang Guoyu was further identified as the top performer.

“Fang Guoyu used AI [原文如此] A well-thought-out strategy eventually defeated it by a narrow margin, but in the ensuing session, AI Fang Guoyu used the same strategy to defeat him,” the Global Times article continued.

Of course, it’s worth noting that this is all according to Chinese state media. Regardless of how long it’s been in Du’s brigade, it’s unclear how widely the PLAAF will use the technology in simulated training or any other application, and how aggressively they might pursue its continued development.

At the same time, it is interesting to note that the PLA Daily itself, not to mention the Global Times, which is affiliated with the Communist Party’s official newspaper People’s Daily, has chosen to highlight an apparently top PLA Air Force pilot with advanced AI in simulated battles, No matter how accurate the description of the capabilities of this simulated opponent is.

Nonetheless, the types of technologies and their capabilities described in the Global Times story are hardly beyond the possible scope of the state of AI technology known to the public. In fact, we were told that the simulated AI-powered adversary the PLA Air Force uses to train its pilots, and Fang Guoyu’s specific experience, sounded very similar to the above-mentioned public broadcast of the AlphaDogfight Trials, a report conducted last year by the U.S. Defense Advanced Research Projects Agency (DARPA). DARPA) LED initiative.

At the conclusion of the three-day event in August 2020, a U.S. Air Force F-16 fighter pilot from the District of Columbia Air National Guard was defeated five times in a row in one-on-one mock air combat. Notably, pilots and experts subsequently raised questions about the validity and applicability of the results of these trials to real-world air combat discussions.

“This (the ability of artificial intelligence and its learning ability) forces pilots to develop more and more innovative tactics and make breakthroughs to win these simulations,” according to the Global Times. “Simulation training can improve training efficiency, save costs, and reduce flight risks. With the rapid development of science and technology, the use of simulation training has become the common goal of the world’s major military powers.”

Broadly speaking, these are potential benefits of integrating artificial intelligence and machine learning into military training regimens. For example, the U.S. military is also increasingly exploring the use of artificial intelligence and machine learning techniques to help improve and save on air combat and other types of training. A particularly notable example is the work of a company called Red 6 in partnership with the U.S. Air Force to develop an augmented reality system that enables pilots of real-world jets to face fully virtual adversaries. You can read more about Red 6 developments that the company hopes will apply to future ground training as well in these past Warzone articles.

Chinese pilots battle AI opponents in simulated dogfights

A U.S. Air Force pilot wears the Red 6’s augmented reality headset on his helmet

In addition to using the technology in training, the report illustrates the PLA’s growing interest in AI for wider applications and what China has already done in this area. Algorithms that can “try first”, virtual adversaries on simulators are likely to be stepping stones to human beings able to operate all levels of real-world unmanned platforms, including fully autonomous unmanned combat air vehicles (UCAVs), another area of ​​development , in which China will continue to make great strides. The Global Times article pointed out that artificial intelligence and machine learning technologies can be applied to manned aircraft to improve efficiency and reduce workload, including helping decision-making in actual combat.

Again, the PLA is not alone in all this. The AlphaDogfight trial is an adjacent effort of a program called Air Combat Evolution (ACE), which more broadly describes the dogfighting capability of autonomous unmanned aircraft as “a gateway to non-linear combat autonomy.” In March, DARPA announced that AI-controlled simulated F-16s were working in pairs in a virtual training environment and hoped to demonstrate the technology on a real small drone later this year. The goal is to integrate the technology into an improved full-size jet trainer by 2023.

Chinese pilots battle AI opponents in simulated dogfights

A DARPA briefing slide showing visually how DARPA envisions the Air Combat Evolution program leading to more advanced air combat autonomous development

The U.S. Air Force is also advancing its Skyborg program, which is developing an artificial intelligence-driven system that it hopes will be able to operate “loyal wingman” drones that work with manned platforms and UCAVs. Some of the technology can also be applied to manned aircraft. An initial version of Skyborg’s “computer brain” underwent its first flight test earlier this year.

Separately, the U.S. Air Force has been preparing a planned demonstration, currently expected in 2024, that could see manned fighter jets engage in real-life dogfights with autonomous drones. This is just some of the work being done around AI and machine learning across the U.S. military and similar developments are taking place in other militaries around the world, as well.

In the United States, interest in these technologies has been particularly high in recent years, due in large part to developments in China. A report released earlier this year by the U.S. government’s National Security Council on Artificial Intelligence put it bluntly that, for now, “the United States is not prepared to defend or compete in the age of artificial intelligence.”

“China’s plans, resources and progress should be of concern to all Americans,” it added. “It’s an AI peer in many fields and an AI leader in some applications.”

So while it’s hard to say what the exact capabilities of the AI ​​that PLAAF pilots are training in simulators might be, it mirrors developments elsewhere, including in the United States. It also underscores the significant investments the Chinese military has made in the field, striving to become a world leader in the application of artificial intelligence technology.

The Links:   G150XTN06B NL6448BC26-11