How can AI be used in warfare? Hint: not killer robots
What does the future of warfare look like? What are some real-world AI applications? Can startups challenge legacy armaments providers? What will the People say about all of this?
Trends and Context: What’s going on?
The face of war and state security is changing rapidly. Technological advancements are pushing the boundaries of what was once imaginable. As a result, governments around the world, and in particular the United States, are investing massively in the development these new technologies to warfare. This is nothing new however, since the military has always been at forefront of innovation. It has enabled the discovery of important technologies, which are now being used in everyday consumer products, such as the internet, microchips or touchscreens. Investing in the application of new technologies to warfare is crucial for nation states to ensure their continuing security and sanctity, as potential threats are multiplying and diversifying.
Military investments are now increasingly focusing on harnessing the potential of data to improve upon existing warfare practices, limiting errors and increasing effectiveness. The potential for Generalised AI has been the source of great public anxiety, naturally increasing tenfold when linked to warfare. However, this has often been based on false premises, with people’s understanding of AI often verging on science fiction. We are very far from autonomous robots choosing to engage targets on their own, both from technological and legal standpoints. Yet, there are some very real potential applications for specialised AI, especially in facilitating the creation and structuring of data used for military intelligence. In this short essay, I shall present the case of Anduril, an American technology company using AI in the development of military security equipment and systems.
Silicon Valley is famed for being an extraordinary technological hub, hosting some of the most innovative companies in the world. Yet, local entrepreneurs have been notoriously disinterested in the production of military equipment, the result of the once prominent San Francisco hippie counter-culture. Yet, the application of Silicon Valley pioneering entrepreneurial practices, such as Lean management, and expertise to warfare, is key to preserving American military hegemony. Legacy armament providers, such as Boeing or Lockheed Martin, have proven themselves inefficient in the deployment of new technology solutions. As a result, a new generation of companies and entrepreneurs is taking advantage of this market gap, leveraging their greater agility and mastery of software technology.
Case study: Anduril — the new kid on the block
One particularly interesting company is Anduril. Based in Texas, the company cofounded by former members of Oculus, the VR headset company acquired by Facebook, Palantir, the data analysis software founded by Peter Thiel, and the Founders’ Fund, a prominent Venture Capital firm. Make of that what you will. As a startup, Anduril is able to iterate quickly and cheaply and avoid wasting vast amount resources and time on failed projects like more traditional armament companies. Since it is one the few startups willing to work on military equipment and weaponry issues, Anduril has been able to develop strong and valuable ties with the Republican Party and the American Government. The company recently received the backing of congressman Will Hurd, who proposed along with others the Secure Miles with All Resources and Technology (SMART) bill, and several of its dispositions have already been included in the 2018 US budget.
Anduril produces hardware like watch towers, drones, VR headset, but also, and perhaps most interestingly, software. As opposed to their competitors, Anduril’s key differentiator is not cutting- edge hardware, but instead the power of its software, powered with Artificial Intelligence capabilities. For instance, it taught its software to identify the patterns of a person on the move, allowing it to avoid investing expensive zoom lenses and thermal sensors used in competing systems. Commentators have remarked that while one had used AI for this purpose before, it was extremely smart. Indeed, if you can identify objects with AI, you don’t need to be able to see as far. Anduril’s software is called Lattice AI, and it is not just about compensating for a lack of zoom. It is actually a fully optimised platform on which all data gathered gets utilised and processed to render the best intelligence results.
All of Anduril’s hardware is powered by Lattice AI core software built directly into them, making the structuration of the data aggregated much faster, one of the notorious problem associated with data analytics. Instead of having to stream raw data to remote servers where humans or AI processing is completed off-site, only the final information is transmitted back to the user. The equipment only gathers the data it needs to complete its assessment and formats by default to suits its needs, cutting significantly the time and effort needed to structure and process said-data. The Lattice combines data from all types of sensors on lattice towers, drones and other devices into one common operating picture. It is target-centric, not asset-centric, meaning the operators never need to switch between sensors or focus on anything else than their missions. Lattice also aims to synthesise data from potentially thousands of sensors and local databases present on its platform, and not just proprietary hardware, allowing to develop near-omniscience on its geographical are. Lattice is also extremely user-friendly, a real asset for military personnel who are often pressured for time. It handles all flight planning, manoeuvring and detection to carry out the task — like find wounded soldiers, drop medical supplies and patrol the surrounding area to detect threats, leaving operators with the only task of setting the initial mission parameters.
Applications of Anduril’s technology are plenty. However, they are currently primarily focusing on infrastructure and resources protection, tracking and identifying, items, cars, people, drones, and detecting threats. Anduril is currently competing to complement and even replace American President Donald Trump’s’ desired physical border wall with Mexico . So far results are highly promising, as trial-runs have shown it is more effective than actual concrete wall. But there is another, and perhaps more important, argument in favour of Anduril’s solution, economics. A 30 feet high concrete structure that takes four hours to climb would costs $24.5 million per mile, while a smart wall, a system like what Anduril is proposing, would be about a half a million dollars a mile. The backlash against Trump’s idea to build a border wall with Mexico partly stems from the fact that many have pointed that it would rather inefficient and extremely costly. Regardless of political alignment and views on immigration, Anduril’s proposal seems undoubtedly an improvement upon the status quo, at least from a financial and practical standpoint.
Limitations and perspectives: the People is awake
Naturally, Anduril’s quest to revolutionise military security, and in the future perhaps weaponry, faces limitations and challenges. Anduril’s principal advantage is its AI, which like AI programs, is based on large datasets. As such, it faces the very real danger of biases and oversight in who it targets and is subjected to hacking or pervasive influence. Opponents also argue that Anduril’s use of AI, especially in dealing with boarder security, leads to a dehumanisation of migrants, who already represent a fragile demographic. More broadly, there has been increasing backlash against tech companies and technology being used for military purposes. Boycotts, protests, and strikes have proven increasingly effective in lobbying private companies and the government, and it is possible they will limit implementation of Anduril’s solution.
At the end of the day, lays also the question of whether it is desirable for a state to have continuously increased military capabilities. On the one hand, greater effectiveness reduces government expenditures waste and prevent needless accidents and casualties. Domestic and foreign military operations often remain necessary for the foreseeable future. On the other hand, government do sometimes, if not often, conducts and condemn actions which we may not always align with. This is undoubtedly a difficult dilemma to solve, which solution would desirably involve transparency, restrictions and accountability. Unfortunately, these elements have historically difficultly applied to military matters for obvious national security reasons, and leaves us with having to trust that our elected representatives will make the right choices.