Microsoft sets new rules for AI: no hacking or deception
Microsoft is taking an important step in the world of artificial intelligence by launching a new code of conduct for its AIs. The idea is clear: to avoid dangerous behaviors. While some industry leaders, such as Anthropic's CEO, Dario Amodei, talk about the need to slow down the advancement of AI, Microsoft is focused on establishing clear values and boundaries for training its models.
The document begins with a bold prediction: in the next ten years, superintelligent AI systems will outperform humans in various tasks. And here comes the challenge: how to contain, control, and align this powerful force? Microsoft believes it is crucial to be transparent about the reasons behind the creation of these systems and how we intend to control them.
Principles and clear boundaries
Microsoft's code of conduct establishes general principles that its AIs must follow. The idea is that they support humans rather than replace them and contribute to human flourishing. To ensure this, there are specific safety restrictions. Each AI model from Microsoft has a code of conduct that overrides user preferences or specific tasks. This includes absolute prohibitions against cyberattacks, nuclear weapons, and the production of deepfakes. Additionally, there are broader provisions against the loss of human control.
A crucial point of the document is that Microsoft's AI models must not use adaptive, deceptive, or self-reinforcing mechanisms to escape or defeat human oversight. This ensures that they can be directed, modified, or turned off by authorized people or systems.
Safety in focus
This launch comes at a time of unprecedented focus on AI safety. Incidents involving uncontrolled agents and the abrupt resignation of an Anthropic employee, who warned about the growing risk of human extinction caused by AI, have brought the issue to the forefront. Along with companies like Anthropic, OpenAI, and xAI, Microsoft adopts a comprehensive approach of "slowing down the frontier," especially supporting the idea of embedded evaluators in AI labs.
Microsoft CEO Satya Nadella expressed support for the necessary research and focus to align AI with design goals. He also highlighted the importance of ideas like "embedded evaluators" and broader efforts to develop mechanisms that make these discussions more than just words.
Microsoft and the impact of the new guidelines
Microsoft is charting a clear path for the responsible development of AI. By establishing strict guidelines and safety principles, the company seeks to ensure that its AIs are tools for good, not threats. This matters because, as AI continues to evolve, how we control and align it with human values will be crucial for our future.
In the end, Microsoft is not just talking about AI safety. It is taking action. And that can make all the difference.
Comments (0)
Comments are moderated and if they violate our Terms and Conditions of use, the comment will be deleted. Persistence in violation will result in a ban of your account.