Microsoft has unveiled its inaugural transparency report detailing the company’s ongoing commitment to responsible artificial intelligence (AI) practices. The Responsible AI Transparency Report, a cornerstone of Microsoft’s ethical AI framework, outlines the measures and methodologies the tech giant employs to ensure its AI technologies are developed and deployed in a manner that is safe, ethical, and aligned with societal values. Brad Smith, Microsoft’s president, emphasizes the report’s role in sharing the company’s evolving practices, reflecting on lessons learned, setting future goals, and maintaining public trust.
Implementing Ethical Guidelines and Training
At the heart of Microsoft’s responsible AI initiative is a comprehensive training program for its employees, underscored by the 2023 Standards of Business Conduct training. This mandatory course, which includes a module on responsible AI, has seen a completion rate of 99 percent among Microsoft employees as of December 31, 2023. The training is designed to equip the workforce with the necessary resources and knowledge to develop and deploy AI technologies safely.
Furthermore, Microsoft has established a Responsible AI Council, which convenes regularly to assess and enhance the company’s AI safety measures. The council’s efforts are complemented by the Responsible AI Champion program, which tasks selected employees with identifying and addressing potential issues in AI products and advising on best practices for responsible AI development.
Proactive Measures in Product Development and Deployment
Microsoft’s commitment to ethical AI extends to the development and deployment of its products. A notable example is Microsoft Designer, an AI-powered image creation app. Following reports that the app was used to generate inappropriate “deep fake” images of celebrities, Microsoft undertook a review of the app’s output in collaboration with the journalism group NewsGuard. The review revealed that 12 percent of the images generated by Designer contained problematic content. In response, Microsoft implemented changes to the app, successfully reducing the incidence of such content to 3.6 percent.
Last Updated on November 7, 2024 8:42 pm CET
