Digital Detectives Instead of Killer Robots
Digital Detectives Instead of Killer Robots: Artificial Intelligence for Global Security
18. November 2024
On November 2, 2024, CNTR participated in the Berlin Science Week program for the first time. During a one-hour Deep Dive lecture, researchers Dr. Niklas Schörnig and Fabian Unruh and science communicator Laura Bannan-Fischer together explored the role of new technologies in strengthening arms control. The focus particularly laid on the use of artificial intelligence (AI) in the verification of nuclear weapons treaties.
Several thousand visitors took part in the CAMPUS program of the Berlin Science Week on November 1 & 2 at the Museum für Naturkunde Berlin. When the CNTR team took over the stage area in the “Evolution in Action” exhibition hall on Saturday afternoon, the hustle and bustle of the museum and festival visitors was still in full swing. Around 50 interested visitors gathered sitting and standing in the secluded stage area at the end of the exhibition space, behind the four-metre-high and 12-metre-long biodiversity wall – an installation showing around 3,000 animals from all habitats –, while the announcement for the deep-dive talk taking place in a few minutes was already displayed on the screen: “Digital Detectives Instead of Killer Robots: Artificial Intelligence for Global Security”.
“AI can help to reduce trust issues, but it can never replace human intuition and interaction.”
Laura Bannan-Fischer, knowledge transfer officer, opened the event with an introduction to the CNTR research and transfer cluster, followed by an introduction to political science research on the development, characteristics and mechanisms of arms control and the role of verification, i.e. the review of compliance with arms control agreements. Verification is constantly confronted with the balancing act of delivering effective, unambiguous, and conclusive examination results on the one hand, while at the same time maintaining the confidentiality of sensitive military and civilian information on the other. This is why the interdisciplinary CNTR team is researching ways to meet this challenge, among others, with the help of AI.
Built on this theoretical foundation, Dr. Niklas Schörnig's presentation dealt with the versatile application possibilities of new technologies in the field of arms control. In fact, these have a long history and were already discussed by researchers in the 1980s (e.g. Allan M. Din's Arms and Artificial Intelligence from 1987). Niklas Schörnig explained that in the areas of arms control and verification, states today benefit in particular from the ability of AI to analyze large amounts of data, which can reveal correlations and patterns that are not necessarily visible to human observers. AI-supported analysis of large amounts of data is used, for example, in simultaneous translation and text analysis, but also in image analysis, such as the search for infrastructure and structural changes in satellite images. One important conclusion was that AI has great potential to support inspectors in their work and can therefore alleviate trust issues in arms control regimes. However, it cannot completely replace human inspectors, as human intuition and interaction remain essential. It is therefore not a panacea for the challenges of arms control.
Finally, Fabian Unruh introduced the audience to the scientific perspective on nuclear disarmament. His presentation focused on two cases in which AI can be used. On the one hand, AI can support the detection of nuclear weapons tests. Such tests are prohibited for states party to the Comprehensive Nuclear-Test-Ban Treaty (CTBT). AI can be used to evaluate the data, which is gathered, for example, using sound measurements from seismic (earth tremors), hydroacoustic (sound in water) or atmospheric sensors. By identifying the location, cause and strength of an explosion, it is possible to detect whether it resulted from an (illicit) nuclear weapon test. For the AI to be precise and reliable, large amounts of simulation data are required to train the system. The second possible area of application for AI systems is the verification of nuclear stockpiles that could potentially be used for nuclear weapons. For this approach, the highly radioactive waste from nuclear reactors is measured. The determined composition of the fission products – the remains of nuclear fission – is analyzed with the help of AI, which identifies the amount of plutonium produced, for example.
Interdisciplinary research for a deeper understanding of global security issues
The CNTR Deep Dive lecture invited the audience to approach this year's guiding question of the Berlin Science Week "What is our common ground?" by highlighting an innovative interface between peace research, technology, and physics, presenting new perspectives on the use of artificial intelligence, and thus establishing a deeper understanding of global security issues and scientific culture among the audience. The importance of such meeting formats was demonstrated in the second part of the event, a half-hour Q&A. The audience participated with questions on various aspects of arms control, such as the effectiveness of future treaties and verification mechanisms or the role of Iran, but also on the research process, such as the development of research questions and knowledge transfer at CNTR.
