The Norwegian “Teknologirådet” Wrote Thursday 26.10. 2023
More and more people are advocating for a global organization for the governance of artificial intelligence. Perhaps the International Atomic Energy Agency or the UN Climate Panel can be used as a template.
In May, more than 350 researchers and technology leaders signed this open letter – in one sentence:
“Mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war.”
The message was clear – managing the risks associated with the development and use of artificial intelligence must be higher on the global priority list. This one-sentence letter followed another open letter from March in which several thousand experts and academics called for a six-month pause in the training of all artificial intelligence systems more powerful than GPT-4. Both letters express a fundamental concern for the consequences of using artificial intelligence, if we do not manage this in a good way.
Several people are now calling for a form of global management of technology development, and there are ongoing discussions internationally about what kind of management model can and should be the basis.
Four models for global governance
In July, researchers from Google’s DeepMind, OpenAI, Harvard, Colombia University and several other academic institutions proposed four models for global governance of artificial intelligence:
• An international institution for knowledge compilation, which can shed light on the opportunities and risks of artificial intelligence, inspired by the UN climate panel. Such an expert group can contribute to increased awareness and understanding of a number of ongoing and upcoming issues for the development and use of artificial intelligence. This can be compared with the UN climate panel (IPCC).
• A global organization for managing artificial intelligence, similar to the International Atomic Energy Agency (IAEA) or the International Civil Aviation Organization (ICAO). Such an organization can ensure a more coordinated global effort to manage and minimize risks from advanced systems for artificial intelligence, as well as assist in the development of norms and standards for the technology.
• An international public-private collaboration, inspired by, for example, the GAVI foundation where government and private actors collaborate to increase access to vaccines in developing countries. The purpose will be to assist several countries in reaping the benefits of breakthroughs in artificial intelligence.
• An international research and development collaboration, inspired by the European Organization for Nuclear Research (CERN). Such cooperation can include member states, researchers and developers of artificial intelligence systems. The experts will have access to computing power and advanced AI technology, in order to be able to develop technical solutions to manage (and minimize) risk.
Four models for global governance
In July, researchers from Google’s DeepMind, OpenAI, Harvard, Colombia University and several other academic institutions proposed four models for global governance of artificial intelligence:
• An international institution for knowledge compilation, which can shed light on the opportunities and risks of artificial intelligence, inspired by the UN climate panel. Such an expert group can contribute to increased awareness and understanding of a number of ongoing and upcoming issues for the development and use of artificial intelligence. This can be compared with the UN climate panel (IPCC).
• A global organization for managing artificial intelligence, similar to the International Atomic Energy Agency (IAEA) or the International Civil Aviation Organization (ICAO). Such an organization can ensure a more coordinated global effort to manage and minimize risks from advanced systems for artificial intelligence, as well as assist in the development of norms and standards for the technology.
• An international public-private collaboration, inspired by, for example, the GAVI foundation where government and private actors collaborate to increase access to vaccines in developing countries. The purpose will be to assist several countries in reaping the benefits of breakthroughs in artificial intelligence.
• An international research and development collaboration, inspired by the European Organization for Nuclear Research (CERN). Such cooperation can include member states, researchers and developers of artificial intelligence systems. The experts will have access to computing power and advanced AI technology, in order to be able to develop technical solutions to manage (and minimize) risk.
The debate is intensifying
More and more actors are joining the debate on global governance.
UN: UN Secretary-General António Guterres believes that a UN agency for artificial intelligence will be able to strengthen global cooperation on the technology. The theme and structure of such an agency has not been clarified. In order to develop a proposal for such an agency, the UN has announced a new, independent panel of experts, where UN member states have nominated experts. This so-called High-Level Advisory Body on Artificial Intelligence was launched in October. They will analyze the possibilities for international management of artificial intelligence, and the first interim report must be completed by the end of the year.
EU: Management of artificial intelligence has also been on the agenda in the EU. A voluntary AI pact consisting of democratic principles for the governance of artificial intelligence, a rulebook for responsible artificial intelligence and a global panel of experts are among the proposals that have been discussed. In her “State of the European Union” speech, European Commission President Ursula von der Leyen highlighted the need for a global approach to understanding the importance of artificial intelligence in our societies. She proposed a new body for artificial intelligence on risks and opportunities for humanity, something similar to the UN climate panel.
China: Recently, China has also announced its proposal for global governance. President Xi Jinping launched The Global AI Governance Initiative at the Belt and Road 2023 Forum (a Chinese initiative to develop infrastructure in more than 150 countries). The aim is to ensure “a human-centred” development of the technology, also outside China’s borders, and promote international standards for testing and risk management.
Technology companies: The big technology companies have also signed up for the debate. OpenAI, the company behind ChatGPT, wants a global AI body, modeled after the International Atomic Energy Agency. They envision such a body being able to inspect systems, require audits and test for compliance with security standards. Another initiative comes from Eric Schmidt (former head of Google) and Mustafa Suleyman (co-founder of Inflection and DeepMind) who calls for an independent, expert-led body with a mandate to inform the authorities about the status of AI, as well as analyze what is to come. Their inspiration is the UN climate panel.
The British are on the ball
With its new role outside the EU, the UK has marked itself out as a new player in the field, with slightly different approaches to the EU, USA and China. This also applies to initiatives for global governance and cooperation. The British government is expected to announce an international advisory group for artificial intelligence at the conference AI Safety Summit in Bletchley Park in early November. This may also turn out to be based on a model inspired by the UN climate panel. It remains to be seen which global governance models for artificial intelligence will make it all the way. Many questions are still unresolved, for example how many global initiatives are needed, and who will take the lead in the work going forward? Nor will it be an easy task to ensure compliance between all the existing guidelines, principles and legislation for artificial intelligence that is under development, both nationally and internationally