Fellow Jeffrey Snover reflects on the STAMP Safety Design Workshop at MIT in a blog post, arguing that 'AI safety' treats the technology as a system, rather than as a component of a system. "Stop waiting for AI to be safe. Start building systems that are."