The Chief Digital and Artificial Intelligence Office of the Defense Department plans to award contracts up to $200 million each to Anthropic, Google, OpenAI, and xAI for developing agentic AI workflows aimed at enhancing national security capabilities. While OpenAI employs reinforcement learning to align its AI models with human-centric values, Anthropic’s Claude uses a published constitution based on principles from the Universal Declaration of Human Rights. This approach seeks to minimize harmful outputs and includes perspectives from less industrialized nations. However, concerns arise regarding xAI’s transparency, as it lacks comprehensive documentation of its guiding principles, leaving critics wary. Experts suggest that while the government’s investment in AI, constituting less than 0.1% of the defense budget, could enhance operations, careful evaluation of each model’s alignment with national values is crucial to ensure efficacy and alignment with security needs.
Source link
Share
Read more