Microsoft Enlists Code of Conduct for AI Models Amid Safety Concerns
Microsoft has released an AI code of conduct aimed at guiding its models away from dangerous behavior. The document acknowledges that superintelligent AI systems will surpass human performance in most tasks within the next decade, posing a significant challenge to humanity.
The code of conduct outlines general principles and specific safety constraints meant to implement those principles. It emphasizes supporting humans rather than replacing them and accelerating human flourishing. Additionally, it includes 'absolute constraints' forbidding cyberattacks, nuclear weapons, or deepfake production.
Microsoft's system also ensures that each model has an overarching code of conduct that overrides individual user preferences or specific tasks. This approach is part of the company's focus on AI safety, which has been driven by a string of rogue-agent incidents and concerns about human extinction.