As the second session of the UN’s Group of Governmental Experts convened last month on Lethal Autonomous Weapons Systems to discuss measures relating to the normative and operational framework for emerging technologies, we interviewed Luís Campani. Luís is a Brazilian researcher with an interdisciplinary focus on both International Relations and Software Development, bridging the gap between the technical aspects and policy considerations involved in regulating technology-intensive activities. He is currently serving as a Research Affiliate at the InterAgency Institute, a digital think-tank. He is currently writing a paper on the debate surrounding Lethal Autonomous Weapons Systems (LAWS) in both Spanish and Portuguese. During his time at the InterAgency Institute, Luís has already taken part in the organization of two events on LAWS: “Conference on Autonomous Lethal Weapons Systems, Artificial Intelligence, and Lusophony” and the “1st Interparliamentary Debate on AI Applied to Defence in Iberophone Countries”, facilitating global dialogue and exchange. Representing InterAgency, Luís has engaged in conversations on LAWS in different forums, such as the Luxembourg Autonomous Weapons Systems Conference and the Group of Governmental Experts for the Conventional on Certain Conventional Weapons (CCW/GGE), showcasing the think tank’s dedication to the subject matter.
Q: Could you explain briefly the distinction between automated and autonomous weapons systems?
During the Luxembourg Autonomous Weapons Systems conference, the NATO representative drew a parallel between the aforementioned distinction and the concepts of automated and autonomous weapons systems. Those concepts are not consensual in their definition and in their policy implications. Those concepts relate roughly to online and offline models in the field of Machine Learning.
In Machine Learning, offline learning refers to a pre-trained model that has been developed using a dataset and then deployed to perform specific tasks. The model’s training is static, and it operates based on the knowledge it has already acquired. On the other hand, an online model is one that undergoes frequent or continuous training, updating its knowledge and adapting to new information as it becomes available.
Q: Do you believe a legally binding instrument on autonomous weapons is possible given states’ reluctance to get behind in the arms race? Will it end up being like nuclear weapons, where agreements merely serve to preserve the hegemony of the nuclear haves?
Over the past few years, a notable inefficiency has emerged within the existing policy-making framework pertaining to lethal autonomous weapons systems (LAWS). This inefficiency has prompted the need for a reevaluation of the approach taken thus far. However, it is essential to acknowledge that there are still untapped potential forums that remain unexplored, which could potentially provide alternative avenues for addressing the regulation of LAWS.
In this context, it is worth noting that when comparing LAWS to other weapons technologies, such as nuclear weapons, there is a crucial distinction. Unlike the case of nuclear weapons, where regulation was implemented reactively following their development and use, with LAWS, we still possess an opportunity to proactively establish regulations before their widespread deployment. This proactive stance allows us to adopt a preventative approach, aimed at addressing potential risks and challenges associated with the use of LAWS before they become more pervasive and difficult to control.
Therefore, despite the inefficiencies observed in the current policy-making framework, it is crucial to recognize the window of opportunity that exists for regulating LAWS in a prophylactic manner. By actively engaging in discussions and exploring untapped forums, we can collectively work towards establishing comprehensive and effective regulations that prioritize human rights, ethical considerations, and ensure the responsible development and use of autonomous weapons systems.
Q: Even if states don’t use AI in battle, they may use it for more ‘civilian’ uses around the battlefield in which they won’t want their servicemen getting hurt (such as going to check out an IED on the side of the road), how will states be able to prevent then the ‘mission creep’ of this eventually being used on the battlefield?
One crucial distinction that sets apart AI surveillance equipment from lethal autonomous weapons systems (LAWS) lies in their respective capabilities for automated targeting and engagement. While AI surveillance equipment is designed for monitoring and gathering information, LAWS possess the ability to autonomously select and engage targets. This distinction raises significant concerns regarding the potential applications of LAWS beyond the traditional context of armed conflict.
Q: In your organisation’s policy brief on the limits to AI, it looks at the ethical dimension of killer robots, could you explain briefly this position and respond to those who argue that they don’t care if they’re killed in battle by a human or a machine just so long as they don’t die needlessly?
In the context of being targeted by a lethal autonomous weapon system (LAWS), it is crucial to acknowledge the profound implications it has on the perception of human value and the concept of dehumanization. When an individual becomes the target of such a machine, they are reduced to a collection of HEX codes that are identified and processed through computer vision algorithms. This process of digital dehumanization occurs when the essence of humanity is stripped away, and individuals are reduced to mere data points. The utilization of this data to inflict harm upon a human being further exacerbates the ethical concerns associated with LAWS.
Digital dehumanization represents a fundamental challenge in the development and deployment of autonomous weapons systems. By reducing individuals to data points, there is a risk of diminishing their inherent worth, dignity, and rights. This devaluation of human life can have far-reaching consequences, not only in the context of armed conflict but also in various other domains where autonomous systems are employed.
Q: As your report mentions, AI technology is already a reality, and as ChatGPT shows, we are very quick to welcome AI if it makes our lives easier, how will that translate on the battlefield when humans will inevitably want machines to take increasingly complex decisions they don’t want to have to make?
The rapid advancement and widespread adoption of AI technology, as evident in the capabilities showcased by ChatGPT, highlight society’s inclination to embrace AI when it simplifies and enhances our daily lives. However, the translation of this acceptance onto the battlefield, where humans may increasingly rely on machines to make complex decisions they prefer not to face, raises profound questions.
There is a need for a thoughtful examination of what specific tasks and responsibilities we are comfortable entrusting to AI systems. Do we envision AI primarily taking on the heavy-lifting tasks, such as data analysis and logistical operations, while humans retain ultimate decision-making authority? Or are we open to AI systems assuming a more significant role in the decision-making process, potentially even determining targets and engaging in combat?
