2023.05.24

Key Points

OpenAI’s CEO, Sam Altman, along with President Greg Brockman and Chief Scientist Ilya Sutskever, emphasize the urgent need for robust regulation and strategic planning to address the potential risks associated with superintelligent AI.

OpenAI proposes three key pillars: coordination and regulation among leading efforts, the establishment of an international authority to oversee superintelligence, and the development of technical safety measures. They advocate for public oversight, democratic participation, and user control over AI behavior.

Coordination and Regulation

To ensure the safe development and integration of superintelligence, there is a pressing need for coordination among leading efforts and the establishment of international regulations. Governments or a new organization could play a pivotal role in overseeing the development and setting limits on the rate of AI capability growth. Holding individual companies accountable for responsible actions is also crucial.

International Authority

Similar to the International Atomic Energy Agency (IAEA) overseeing nuclear activities, a governing body must be established to regulate superintelligence efforts surpassing a certain capability threshold. This authority would conduct inspections, enforce safety standards, conduct audits, restrict deployment and security levels, and monitor compute and energy usage to mitigate risks effectively.

Technical Safety Measures

Developing the technical capability to ensure the safety of superintelligence is an ongoing research endeavor. OpenAI and other organizations are dedicating significant effort to address this challenge and prevent potential dangers associated with the unprecedented power of superintelligence.

It is important to strike a balance in regulating AI systems. While allowing companies and open-source projects to develop models below a significant capability threshold without burdensome regulation, the focus should primarily center on regulating systems with power beyond any technology created so far. Existing systems pose risks that can be managed comparably to other Internet technologies.

Public oversight plays a crucial role in the governance and deployment of powerful AI systems. OpenAI advocates for strong democratic participation and input from people worldwide to establish boundaries and defaults for AI systems. OpenAI plans to experiment with mechanisms that allow users to exert control over AI behavior while operating within broad parameters.

OpenAI recognizes two fundamental reasons for pursuing superintelligence. Firstly, it holds immense potential to significantly improve various aspects of society, including education, creativity, and productivity. Secondly, preventing its creation presents substantial challenges due to decreasing costs, increasing actors involved, and its alignment with technological progress. Stopping superintelligence would require drastic measures that might not be feasible. Thus, ensuring the safe development of superintelligence becomes paramount.

As the advent of superintelligent AI draws nearer, global coordination, stringent regulation, and proactive safety measures are imperative to harness its benefits while mitigating risks. OpenAI urges stakeholders to address these concerns promptly to shape a future where superintelligence enhances society and safeguards humanity’s well-being.