Technology from the civilian sphere is widely regarded as an enabler of Lethal Autonomous Weapons Systems. Silicon Valley tech companies, the thinking goes, are far more advanced than government research labs in developing dual-use technologies such as AI and robotics. They are therefore increasingly important to the development of military technologies, but this is contingent on their cooperation. In this paper for Global Policy, Maaike Verbruggen articulates the general dynamics of how this plays out (or fails to, as the case may be).
Civil-military cooperation in researching and developing technologies is seldom straightforward. A range of economic, cultural and practical difficulties hinder the ability of militaries to procure civilian technologies for defence purposes. For many companies, governments are simply not spending enough to warrant the energy and resources required to navigate their bureaucratic procurement processes. As Google’s Project Maven has shown, even when the management of companies are willing to work with defence actors, there is no guarantee their workers will be so compliant. Similarly, starkly different business practices and social networks between these domains may further inhibit cooperation.
Even when these challenges are overcome, the work of then adapting these technologies to the military domain is easier said than done. As Verbruggen notes: “one cannot take an already developed programme, change the names of a few variables and add a machine gun to it.” It goes without saying that there are drastic differences in the operational environment of a war zone compared to an ordinary, civilian setting. Gathering and labelling the data required for training algorithms is a challenging endeavour for any machine-learning model. In the case of LAWS, this is exacerbated by the fact warfare is characterised more by its unpredictability rather than by set patterns and routines.
Verbruggen is careful to note, however, that cooperation (and conflict) between military and civilian spheres over these technologies can vary dramatically. In Israel, where mandatory conscription means 90% of tech workers are veterans and there is consequently a common understanding of operational needs, Verbruggen notes that cooperation between these sectors is relatively straightforward. Similarly, the blurred public-private distinction in China renders its dynamics on this issue very different to the US, where the vast majority of research has focused on so far.
Verbruggen’s contribution to the fast-growing body of literature on LAWS development is a useful guide to navigating this under-theorised phenomenon. Perhaps most promising is her optimism that there is nothing preordained about cooperation in this field. In her words, “effective governance is thus within reach.” How we get there remains to be seen.