What AI Models for War Actually Look Like (www.wired.com)

Smack Technologies has announced a $32 million funding round to develop advanced AI models tailored for military applications, boasting capabilities that could surpass existing models like Claude. Unlike other AI firms such as Anthropic, which express concerns over military use, Smack's leadership, composed of former military personnel, sees value in integrating AI into mission planning and execution while emphasizing ethical usage by trained personnel. Their approach involves training models through simulated war games, using trial and error to identify optimal strategies, akin to Google's training of AlphaGo. The significance of Smack's developments comes at a time when military applications of AI are increasingly scrutinized, especially following contentious discussions between the Department of Defense and AI companies regarding contract terms and the ethical implications of autonomous systems. While there are ongoing concerns about the reliability of AI in high-stakes military scenarios—particularly after studies suggesting large language models could exacerbate conflict—Smack's models aim to streamline mission planning processes traditionally reliant on manual methods. This shift could enhance decision-making speed, crucial in potential engagements with major powers like Russia or China, though experts call for strict governance to mitigate the risks of deploying AI in warfare.