We Cannot Trust AI With Control Of Our Bombs
According to Professor Michael T. Klare, AI has the potential to replace human decision-making in military operations, including the use of weapons of war such as nuclear arms. In the words of Klare, AI's propensity for errors and unpredictable "hallucinations" poses a catastrophic risk if it is given control over weapons, as shown by the potential of machines fighting each other and causing human extinction. Major powers like the US, Russia, and China are rapidly developing AI-powered automated battlefield decision-making systems, which could dictate combat strategies, issue attack orders, and even control nuclear weapons. The development of such systems, like the Advanced Battlefield Management System (ABMS) and the Joint All-Domain Command-and-Control System (JADC2), raises concerns about the potential for AI-triggered conflicts, with opposing AI systems leading to unintended and devastating consequences. Despite these risks, there are currently no effective measures in place to prevent such catastrophic outcomes, highlighting the urgent need for global discussions and safeguards to control the weaponization of AI.
In the near future, AI-controlled systems could determine when and how military forces engage, escalating conflicts and potentially leading to global catastrophe. The convergence of AI technology with military decision-making raises significant ethical, security, and existential concerns that must be addressed to prevent unintended and devastating consequences.