Driving Factors of Autonomous Military AI

Global powers, especially those willing and able to test these systems in war zones around the world, will drive the development of autonomous military artificial intelligence (AI) systems. This will make the technology less useful for smaller, less-resourced countries at first.

Since Russia's attack in February 2022, Ukraine has been used to test a number of self-driving military AI programs. This happened because Western companies worked harder to help Kyiv and because the battlefield in Ukraine gives us access to important military data that can help us improve train systems. There are U.S. and European companies that focus on self-driving drones and more specialized companies that look at data from battlefields to help the military make decisions, identify faces, and do other things.

Russia, which doesn't have a very powerful AI industry, is said to have used AI-powered drones in Ukraine, but it's not clear what Russia's real AI weaponry capabilities are.

• The U.S. data analytics company Palantir has done important work in Ukraine and even opened an office there. Palantir is a government contractor that is known for giving software to U.S. Immigration and Customs Enforcement (ICE), and the CIA was one of the company's early investors. Palantir's software uses AI to look at open-source data, drone footage, satellite images, and reports from the ground to give the military choices. In Ukraine, this has been used for targeting as well as gathering information, clearing land mines, and helping to move refugees.
• The US and UK are working together to give Ukraine thousands of AI-powered swarm drones that could attack Russian targets at the same time.
• Developers from Ukraine say that their Saker Scout drones can find 64 different kinds of Russian military items, identify them, and attack them without a person controlling them.

The Current Use Case

Tech companies are trying to use Ukraine as a real-life testing ground. At the same time, many governments around the world have also been trying to add AI to their forces for self-driving systems.

There is a race between China and the US to make self-driving cars and weapons, as well as AI-powered subs and other vehicles. However, many of these projects are still secret. The US Department of Defense (DoD) has put a lot of money into Project Maven, which is an advanced monitoring system that will let the military find targets automatically. Maven is currently only being used to look at satellite images and other data in order to find possible targets. It is not being used to confirm the targets or automatically launch weapons against them.

More recently, on February 21, the DoD confirmed the deployment of Combined Joint All Domain Command and Control (CJADC2) to improve military operations' battlefield awareness. This shows that the U.S. military has been working hard over the past few years to improve its AI automation skills.

• As China's military danger grows, Australia has put money into making advanced unmanned submarines powered by AI. These are called GhostSharks, and they will be delivered by 2025. Anduril, an AI company, wants to build a factory in Australia so that these submarines can be used "on a large scale." The company also said it wants to make the same type of submarine for the US and its allies, such as Japan, Singapore, South Korea, the UK, and other European countries. These are all well-resourced countries with very developed tech sectors.
• China said in 2023 that it had made an AI drone that could beat a human-controlled unmanned aerial vehicle (UAV) in a battle in the sky.

Military AI Competition Among the Majors

Over the next ten years, the most powerful militaries in the West and China will lead the way in developing autonomous weapons systems. Other countries will be limited by their inability to get access to good training data and AI for the specialization of autonomous systems.

Strong military forces in the West, like the US and UK, as well as China and Israel, are all working quickly to develop AI for use in everyday life. These countries will be at the forefront of using cutting-edge technology for military reasons because they can take advantage of the growing commercial AI talent and development pipeline. This will help determine how the technology will be used by militaries in the future. Long term, these forward-thinking nations will be able to train systems for general use, like predicting weather and ice patterns, which will help they plan their defense. This means that systems will need to be retrained to work well in different places, like deserts, instead of the hilly areas where they were first trained.

Other armies will eventually be able to use them, especially since tech companies in the West are spending money to try new projects in new conflicts. Autonomous systems for the air and sea, in particular, are likely to be some of the first to spread because they can be used in more places and are easier to adapt to different terrains than autonomous systems for the land. But countries that don't have as many resources will still find it hard to train and test AI that works well for self-driving systems at first because of multilateral or Western export controls and a lack of access to technology experts, computing power, and good data for the job.

In fact, many countries that aren't very creative might not have the money or technology to do these kinds of projects, especially when it comes to autonomous land systems. Professionals working on AI development are also more likely to leave for more innovative countries in the West and Asia in order to get better education and skills and move up in their careers. This makes it harder for smaller, less-resourced countries to grow their tech industries and take advantage of the promise of AI innovation.

• Some countries, like Iran, might want to create self-driving AI defense apps that they can use themselves or sell to other countries. But these kinds of systems would probably only be able to learn from testing data from Iran's specific terrain, weather trends, and altitude. This would make them less useful in places like Russia where there are a lot more trees. Iran and other countries with few resources may not think it is worth the time, effort, and money to retrain autonomous military AI systems in settings that are different from the ones they were trained on the first time. This is because the process is long and expensive.
• The first military uses of AI for self-driving systems will also set the stage for how the technology will be regulated around the world. This is mostly going to be a problem for Western countries that follow a set of globally agreed-upon rules. However, it will still affect how autonomous AI is used in the military, and there may be limits on how the technology can be shared.

The Brave New AI World

Western tech companies will lead the way in developing self-driving military AI systems, especially in countries that are at war. This could help less-resourceful countries' tech sectors, but it will also put the companies' reputations at risk when they test and build these technologies in conflict zones.

For instance, the battle-tested AI used in Ukraine has offered advances in strategic warfare apps. Palantir is using AI that was taught on data from the battlefield to help make it easier for tanks and artillery in Ukraine to hit their targets. This means that AI-enhanced targeting features are being designed to work only in Ukraine. They won't be able to be used on autonomous systems in other places, since the things that make targeting better in Ukraine might not work in other places.

Tech companies may use other new conflicts around the world in the same way to test data for new military AI uses. Companies like these will also have a reason to help warring nations get access to real battlefield data so that they can improve their training systems, as long as international laws allow it. In countries that don't have the technological know-how or money to grow this part of their economies yet, this could help move tech industries forward. But companies that get involved in many conflicts, especially controversial ones like the ongoing war between Israel and Hamas in Gaza, will also risk their reputation. This could cause people to stop using their services, the media to focus on them, and them to lose important contracts that aren't related to the military.

So, self-driving AI defense systems won't work as well in other settings, and countries with fewer resources will have a hard time retraining their people to use them.

If a country with few resources decides to spend money on building autonomous AI defense systems, it might share the software with countries nearby that have similar environments. For instance, countries can use improvements in transfer learning and domain adaptation methods to make apps work better in new settings, but they would need to have access to top-notch technical experts and a tech sector that is well-developed.

Western companies in Ukraine have been given incentives to make apps that protect democracy and give them direct access to training data from the battlefield for their own reasons.

• The way AI-powered drones are built and used is very different from how remote-controlled drones are used. A lot of drone users in Ukraine can take out much more advanced (and expensive) Russian weapons with cheap, plastic, commercial drones. Because sensors and algorithms can turn what people say into exact orders for the drone to follow, drone technology has become very easy for many people to get. AI-powered drones, on the other hand, can fly themselves without any help from a person and need huge amounts of training data to learn how to follow instructions on their own. In the past ten years, remote control technology has spread quickly as it became more widely usable for business reasons after being first used in the military in the 1980s.