Microsoft Emphasizes Humanist Approach to AI Development
Microsoft has released an 11,000-word draft 'Code of Conduct' for its researchers to follow when training and operating AI models. The document outlines a philosophy of 'Humanist AI' and emphasizes that AI should not be designed to imitate consciousness. This stance comes amid calls from leaders in the field to 'pace the frontier' of AI development due to concerns about potential risks, including agentic loss of control.
The company's position is notable as it has recently released its MAI-Thinking-1 model, which was trained on clean data without distillation from third-party models. Microsoft AI CEO Mustafa Suleyman acknowledged that fears about possible loss of control are real, but the company has not explicitly called for a slowdown in development.
The Code of Conduct is currently undergoing a six-week public consultation and is not yet being used for training purposes. The document aims to establish values and behavior for in-house AI models, which may have significant implications for the field as a whole.