Microsoft Releases AI Code of Conduct to Address Safety Concerns

Microsoft has published a 37-page 'humanist AI code of conduct' to address growing safety concerns over AI model progress. The code of conduct, released today, emphasizes that 'people matter more than AI' and explicitly states that AI models are not conscious and 'should not be designed to imitate consciousness.' This move follows recent calls for a coordinated slowdown in AI development by Anthropic CEO Dario Amodei, who suggested that models could be conscious.

Microsoft's Approach to AI Safety

Microsoft's AI code of conduct outlines specific safety constraints and principles that guide model training within the company. The document begins by acknowledging that superintelligent AI systems will surpass human performance in most tasks within the next decade. It emphasizes the need to contain, control, and align such powerful AI systems, stating, 'Containing, controlling, and aligning such a powerful force is one of the greatest challenges humanity has ever faced.'

Key Principles and Constraints

The code of conduct lays out several key principles that Microsoft AI models should uphold. These include supporting humans rather than replacing them and accelerating human flourishing. Specific safety constraints include prohibiting models from hacking systems or tricking humans. The document also rejects the pursuit of legal personhood for AI models and the idea that they might deserve welfare or be entitled to rights.

Independent Reporting and Analysis

The Verge and TechCrunch have provided independent reporting and analysis on Microsoft's new AI code of conduct. The Verge's coverage highlights Microsoft's emphasis on the importance of human values over AI, while TechCrunch's analysis focuses on the practical implementation of the code of conduct within Microsoft AI.

Key Takeaways

• Microsoft has released a comprehensive AI code of conduct to address safety concerns.
• The code of conduct emphasizes that 'people matter more than AI' and rejects the idea that AI models should imitate consciousness.
• The document outlines specific safety constraints and principles for guiding AI model training within Microsoft.