Exploring institutions for global AI governance

A new white paper is looking at the models and functions of international institutions that can help manage opportunities and relieve advanced AI risks

The growing awareness of the global impact of advanced artificial intelligence (AI) has inspired general discussions about the need for international governance structures to help manage opportunities and reduce the risks involved.

There have been many discussions on the likeness of ICAO (International Civil Aviation Organization) in civil aviation; CERN (European Nuclear Research Organization) in particle physics; The International Atomic Energy Agency (International Atomic Energy Agency) in nuclear technology; And international governmental organizations with stakeholders in many other areas. However, although metaphors can be a useful start, the techniques arising from artificial intelligence will be unlike aviation, particle physics or nuclear technology.

To succeed in the governance of artificial intelligence, we need a better understanding:

1. What are the specific benefits and risks we need to manage internationally.
2. What is the governance that requires these benefits and risks.
3. What organizations can provide these functions better.

Our last paper, with collaborators from Oxford University, University of Montreal, University of Toronto, University of Colombia, Harvard University, Stanford University, and Openai, deal with these questions and achieve how international institutions can help manage the global influence to develop the AI, and make sure of the benefits of AI.

The decisive role of international and multilateral institutions

Access to some technology of artificial intelligence can greatly enhance prosperity and stability, but the benefits of these technologies may not be evenly distributed to the greatest needs of the representative societies an incomplete representation or the developing world. Insufficient access to Internet services, computer capacity, training or experience of machine learning may also prevent certain groups from taking advantage of fully from progress in artificial intelligence.

International cooperation can help address these issues by encouraging organizations to develop systems and applications that meet the needs of disadvantaged societies, and by improving education, infrastructure and economic obstacles that face such societies that benefit fully from artificial intelligence technology.

In addition, international efforts may be necessary to manage the risks posed by strong artificial intelligence capabilities. Without adequate guarantees, some of these capabilities – such as developing automatic software, chemistry and artificial biological research, and generating text and video can be used to cause harm. Advanced artificial intelligence systems may also fail in difficult ways, creating an accident risk with possible international consequences if technology is not published responsibly.

International and multi -stakeholder institutions can help develop artificial intelligence development protocols that reduce these risks. For example, they may facilitate the global consensus on the threats posed by the various artificial intelligence capabilities of society, and to set international standards on identifying models and treating them with serious capabilities. International Safety Research will enhance our ability to make systems reliable and flexible for misuse.

Finally, in cases where countries have incentives (for example, derived from economic competition) to undermine organizational obligations to each other, international institutions may help support and stimulate best practices and even monitor compliance with standards.

Four possible institutional models

We explore four supplementary institutional models to support global coordination and governance functions:

• International governmental AI Border Committee An international consensus can be built on the opportunities and risks of advanced AI and how it can be managed. This would increase public awareness and understand the prospects and issues of artificial intelligence, contribute to a scientific enlightened account for the use of artificial intelligence and risk mitigating, and to be a source of experience of policy makers.
• Governmental Advanced Advanced Intelligence Governance Organization It can help internationalize and align efforts to address global risks of advanced artificial intelligence systems by setting rules and criteria for governance and helping them. Compliance monitoring functions may also perform any international governance system.
• A AI cooperative borders It can enhance access to advanced artificial intelligence as an international partnership between the public and private sectors. By doing this, this will help benefit from the developed artificial intelligence technology and enhance international access to artificial intelligence technology to achieve safety and governance goals.
• and Artificial intelligence safety project The main researchers and engineers can bring them together by accessing arithmetic resources and advanced artificial intelligence models to research artistic relief of the risks of artificial intelligence. This would enhance the research and development of artificial intelligence by increasing its scope, resources, and coordination.

Operating challenges

Many important open questions remain about the feasibility of these institutional models. For example, the AI ​​Committee will face important scientific challenges due to the lack of certainty about the paths and capabilities of artificial intelligence and limited scientific research on advanced artificial intelligence issues so far.

The rapid rate of progressive intelligence and limited ability in the public sector may make artificial intelligence issues that are difficult for the advanced artificial intelligence governance organization setting standards to keep pace with the scene. The many difficulties of international coordination raises questions about how countries motivate to adopt their standards or accept their observation.

Likewise, many obstacles to societies that fully mock the benefits of advanced artificial intelligence systems (and other technologies) may keep Amnesty International Cooperative to improve their influence. There may also be difficult to manage it between sharing the benefits of artificial intelligence and preventing the spread of dangerous systems.

As for the artificial intelligence safety project, it will be important to carefully consider safety research elements that are done better through cooperation in exchange for individual efforts of companies. Moreover, the project can fight for sufficient access to the most able models to conduct safety research from all the developers concerned.

Looking at the tremendous global opportunities and challenges provided by artificial intelligence systems on the horizon, there is a need for further discussion between governments and other interests of interest about the role of international institutions and how their employees can increase the governance of artificial intelligence and coordination.

We hope this research will contribute to the growth of talks within the international community on ways to ensure advanced artificial intelligence in favor of humanity.