The new White Book is investigating models and functions of international institutions that could help manage possibilities and reduce the risk of advanced AI
The growing awareness of the global influence of advanced artificial intelligence (AI) inspired public discussions on the needs of international management structures to help manage possibilities and reduce risk.
Many discussions were based on analogies of Icao (International Civil Aviation Organization) in civil aviation; CERN (European Nuclear Research Organization) in particle physics; Iaea (International Atomic Energy Agency) in the field of nuclear technology; and intergovernmental and multi -part organizations in many other domains. And yet, although analogies can be a useful beginning, technologies emerging from AI will be different than in aviation, particle physics or nuclear technology.
To succeed in managing AI, we must better understand:
- What specific benefits and risk do we need to manage internationally.
- What management works, these benefits and risk require.
- What organizations can best ensure these functions.
Our latest articlewith colleagues from the University of Oxford, Université de Montréal, University of Toronto, Columbia University, Harvard University, Stanford University and OpenAI, he deals with these questions and research on how international institutions can help in managing the global influence of the development of Frontier Ai and make sure that the benefits community.
A critical role of international and multilateral institutions
Access to some AI technologies can significantly increase prosperity and stability, but the benefits of these technologies may not be evenly distributed or focused on the greatest needs of insufficiently represented communities or a developing world. Insufficient access to internet services, computing force or availability of training or specialist knowledge of machine learning may also prevent some groups from full use of progress in artificial intelligence.
International cooperation can help solve these problems, encouraging organizations to develop systems and applications that meet the needs of underestimated communities and by improving education, infrastructure and economic obstacles for such communities that fully use AI technology.
In addition, international efforts may be necessary to manage the risk created by the powerful possibilities of artificial intelligence. Without adequate security, some of these possibilities – such as automated software development, chemistry and synthetic biology tests, and text and video generation – can be used improperly to cause damage. Advanced AI systems can also fail in a difficult way to predict, causing the risk of accidents with potentially international consequences if the technology is not implemented responsibly.
International and multi -part institutions can help develop AI development and implementation protocols that minimize such risk. For example, they can facilitate the global consensus as to the threats that the various possibilities of AI constitute society, and establish international standards regarding identification and treatment of models with dangerous possibilities. International cooperation on security research would also increase our ability to make systems reliable and resistant to improper use.
Finally, in situations where countries have incentives (e.g. coming out of economic competition) to cut each other, international institutions can help support and encourage the best practices, and even monitor the following standards.
Four potential institutional models
We study four complementary institutional models to support global coordination and management functions:
- Of intergovernmental Commission on the border of AI He could build an international consensus regarding the possibilities and risk of advanced artificial intelligence and the method of managing them. This would increase social awareness and understanding of AI's prospects and problems, would contribute to the scientifically informed description of the use of AI and reduction of risk, and would be a source of specialist knowledge for decision -makers.
- Intergovernmental holder or many times Advanced AI management organization It can help internationalize and adapt efforts to solve the global risk of advanced AI systems, establishing management standards and standards, and helping in implementation. It can also perform compliance monitoring functions for each international management system.
- AND Frontier Ai Collaborative It can promote access to advanced artificial intelligence as an international public-private partnership. In this way, this would help underestimated societies to use the latest AI technology and promote international access to AI technology for security and management purposes.
- Some AI safety project He can gather leading researchers and engineers and provide them with access to computational resources and advanced AI models for research on technical alleviating AI threats. This would be conducive to research and development of AI safety by increasing its scale, resources and coordination.
Operational challenges
There are many important open questions about the profitability of these institutional models. For example, the AI advanced committee will face significant scientific challenges, taking into account the extreme uncertainty about the trajectory and AI abilities, as well as limited scientific research on advanced AI problems.
A quick AI progress rate and limited ability in the public sector in matters related to AI Frontier may also make it difficult to determine the advanced AI management organization that keeps the standards that keep up with the landscape of risk. Many difficulties in international coordination give birth to questions about how countries will be encouraged to accept their standards or accept its monitoring.
Similarly, many obstacles to societies fully use the benefits of advanced AI (and other technologies) systems, they can maintain AI cooperation from the optimization of its impact. It can also be difficult to manage between sharing AI benefits and preventing the spread of dangerous systems.
And for the AI security project, it will be important to carefully consider which elements of security research are best carried out through cooperation compared to the individual efforts of companies. In addition, the project can fight to provide adequate access to the most talented models to conduct security research for all relevant programmers.
Considering the huge global possibilities and challenges presented by AI systems on the horizon, a greater discussion is needed among governments and other stakeholders about the role of international institutions and their functions may be further management and coordination of AI.
We hope that these studies contribute to growing conversations in the international community on how to ensure an advanced AI development for humanity.
