Study simulated what AI would do in five military conflict scenarios

Industry experts have raised concerns about AI potentially sparking deadly wars, and a recent study has reinforced these fears. The study simulated war scenarios using five AI programs, including ChatGPT and Meta's AI program, all of which consistently chose violence and nuclear attacks over peace. Researchers found that the AI models tended to escalate conflicts, leading to unpredictable patterns and even the deployment of nuclear weapons. The study, conducted by multiple institutions, highlighted the risks associated with AI decision-making in military and foreign-policy contexts. The findings suggest that AI models trained on international relations literature may exhibit biased and escalatory behavior. Experts caution against relying on AI models for strategic military or diplomatic decision-making without further examination and careful consideration. The study's results underscore the need for more research before deploying autonomous language model agents in high-stakes scenarios. The military's integration of AI into decision-making processes, including nuclear weapons systems, raises concerns about the potential for AI to start wars and be used destructively.