AI Warfare: No Going Back

AI warfare is here to stay. At the inception of the greatest breakthroughs in the history of computing, we can only speculate how things will go both geopolitically and for humanity as a whole.

The U.S. seeks to develop AI-enabled swarming drones for counter A2AD operations.

On Friday, November 27, 2020, Iran’s top nuclear physicist, Mohsen Fakhrizadeh, drove with his wife from his holiday home on Iran’s Caspian coast to his country house in Absard, near Tehran. Although Iranian authorities warned him to avoid travel due to security issues, Fakhrizadeh insisted on traveling to the capital for an important meeting and to deliver a lecture to his students at Iman Hussein University. The rural road on which he drove his armored Nissan Teana, accompanied by three other armored vehicles, was less busy than usual due to COVID-19 restrictions.

Suddenly, a shot rang out from a pickup truck on the other side of the road. Not recognizing the gunfire and perhaps thinking he had struck something, Fakhrizadeh stopped and stepped out of the vehicle. Then came additional shots, hitting Fakhrizadeh thirteen times and killing him along with a bodyguard who attempted to intervene. Then, the pickup truck exploded. The entire attack, widely believed to have been perpetrated by Israel, occurred in less than three minutes. What’s more, the gun used in the attack was operated entirely remotely via satellite while using artificial intelligence (AI) facial recognition to target the victim specifically. It was one of the world’s first glimpses into the growing trend of harnessing AI for military purposes.

Militaries worldwide are adopting AI technologies to meet the battlefield needs of the future. In this article, we explore what AI technologies are in use or near development today and the possible future of AI-based warfare systems in a post-microchip world.

An aerial map of the site of Fakhrizadeh’s assassination using an AI-enabled machine gun. Source

AI weapons are already here

AI weapons are a growing reality of today’s battlefield, with many examples of AI-based weapons systems already in use or under development. In the U.S., the Pentagon awarded General Atomics a $93.3 million contract to equip one of its MQ-9 Reaper drones with AI technology that would enable it to fly autonomously and identify potential ground targets. This proved problematic when, during a simulated mission, the drone attacked and killed its operator for interfering in the mission—something the U.S. Air Force officially denied (no one was actually harmed). The U.S. Army also recently contracted Palantir to develop a ground intelligence system that employs AI to identify targets beyond soldiers’ line of sight.  

AI-powered autonomous vehicles are becoming the norm. In 2021, the Australian military contracted defense firm GaardTech to supply it with the Jaeger Type-C, an autonomous ground vehicle equipped with an armor-piercing machine gun, sniper rifle, and over 100 pounds of explosives for kamikaze attacks. Designed to attack in swarms, these AI-enabled units can hit targets from multiple angles at once: one demonstration video shows one of these units approaching the target to confirm its destruction. Similarly, the Pentagon’s Defense Advanced Research Projects Agency (DARPA) called for companies to submit proposals for swarming seaborne attack drones, and the U.S. Navy reportedly tested swarming aerial and seaborne drones at its annual UNITAS exercise last July. DARPA specifically claims that swarming drones are to be used in counter anti-access area denial (A2AD) operations – please see my previous articles for a more in-depth exploration of A2AD and its impact on contemporary warfare.

China is also a world leader in AI-enabled drone technology, investing heavily in robotics, swarming, and other technologies. The world’s largest producer of commercial drones, China’s AI attack drones are reportedly becoming more sophisticated. People’s Liberation Army (PLA) researchers claim that a domestically produced AI-enabled Wing Loong 2 attack drone beat a human-operated unit in an aerial battle test. In 2018, Chinese state-owned defense contractor Norinco introduced the Sharp Claw II, an autonomous ground vehicle designed to conduct surveillance, patrol, assault, and transport in dirty, monotonous, and dull combat environments. Another Chinese manufacturer also unveiled the autonomous and amphibious Luwu 8x8 ground vehicle in 2018: amphibious units are likely to play a crucial role in any Taiwan invasion scenario.

AI drone usage is not limited to advanced industrial countries. A 2021 United Nations report found that someone in Libya – likely the Government of National Accord – used a Turkish-made Kargu-2 attack drone with AI capabilities to kill targets. The drone reportedly had no human operator and used facial recognition technology to hit human targets specifically. 

An American MQ-9 Reaper. While being tested with an AI system, one of these reportedly killed its operator in a simulation (no one was actually harmed). Source

A Chinese Luwo 8x8 amphibious autonomous drone. Source

AI chips as a means of containment

AI-enabled devices require microchips uniquely designed for AI to function. Unlike regular chips, AI chips have enhanced parallel processing capabilities, meaning they can do more things simultaneously. Only a handful of companies on earth can manufacture these chips, with Nvidia being the largest, followed by Intel, Alphabet, Apple, IBM, and a few other American companies.

Until 2022, Nvidia supplied roughly 90% of China’s AI chips, but this practically ceased in the fall of that year due to trade restrictions from the U.S. Department of Commerce. Although some Chinese firms such as Tencent claim to have stockpiled enough Nvidia chips to continue developing their AI programs, companies there are now searching for domestic providers. Although domestic firms such as Huawei are developing locally produced alternatives, these chips are less powerful than those available in the U.S. Moreover, Nvidia has introduced a U.S. trade-restriction-friendly alternative to compete with these domestically-made chips. However, reports indicate that Chinese military and government buyers continue to purchase small batches of top-of-the-line Nvidia chips despite export bans. For this reason, the global AI chip market has already become difficult to regulate.

Into the quantum

Moore’s law speculates that the number of transistors on a chip will double roughly every two years. However, with transistors now smaller than bacteria or a virus, we are approaching the limits of how far chip technology can advance as quantum bits of data (qubits) will be able to travel across a chip uncontrollably if transistors get much smaller, rendering the chip inoperable. Instead, computer scientists are harnessing qubits in an entirely new manner called quantum computing. One hundred fifty-eight million times faster than the world’s most advanced chip-based supercomputers, quantum computers are likely the future of AI technology, including AI military technology. However, quantum computers are difficult to build, impossible to mass produce, and highly costly; as such, only a tiny handful of organizations can advance this technology.

Companies in the U.S. and its allies continue to make significant strides in quantum computing. However, some speculate that China is nearing U.S. quantum capabilities or has even reached parity. Scientists now warn that Q-day – when AI-enabled quantum computers can crack any encrypted code – could come as soon as 2025, putting all of the world’s private information and network devices at risk. For this reason, the U.S. military contractors developed quantum-resistant data security measures for autonomous drones. However, it is difficult to tell whether such security measures can effectively evolve to meet the challenge of advanced quantum computers moving forward.  

The most important thing about quantum computing concerning AI is its ability to leapfrog advanced chip technology. In this way, the countries with the most advanced quantum computers and quantum networks will be able to run the most advanced AI-based military systems. If China is as close to the U.S. as some speculate, it could be a geopolitical game-changer.

A quantum computer is reportedly 158 million-times faster than the world’s most advanced supercomputer. Source

Conclusion

It is highly likely that what we have seen to date is only the beginning of a revolution in AI-based military technologies. What’s more, by potentially pairing autonomous artificial intelligence with our greatest advances in computing technology, we cannot possibly foresee the consequences of these developments. One can only hope that safeguards are put in place before it is too late.