Tech Review Explains – Let our writers help you untangle the complicated, messy world technology. Read more about it here.
AI regulation is a hot topic. Since the success of OpenAI’s ChatGPT chatbot, the public has been captivated by wonderment and concern about the powerful AI tools. The potential of generative AI to revolutionize productivity tools and creative assistants has been hailed. They are already causing harm. They have been used as a tool to spread misinformation and could be weaponized for spamming or scamming.
In recent weeks, everyone from US Senators to tech CEOs at G7 meetings has called for stronger AI standards and international guidelines. The good news is? The good news is that policymakers do not have to start at the beginning.
We have analyzed six international attempts at artificial intelligence regulation, outlined the pros and cons for each and assigned a rough score to indicate how influential we believe they are.
A legally-binding AI treaty
The Council of Europe is a 46-country human rights organization. It is currently finalizing a legal treaty on artificial intelligence. Signatories are required to take measures to ensure AI is developed and used in a manner that respects democracy and human rights. The treaty may include moratoriums for technologies such as facial recognition that could pose a threat to human rights.
According to Nathalie Smuha who is a legal scholar at the KU Leuven Faculty of Law and advises the Council, if all goes well, the organization may finish drafting the document by November.
Pros The Council of Europe has many non-European nations, such as the UK, Ukraine and the US. It also includes Canada, Israel and Mexico. Smuha says, "It is a very strong signal."
Con: Each country must ratify and implement the treaty in its national law individually, which can take many years. It's possible that certain countries could opt out of elements they dislike, like strict rules or moratoriums. Smuha says that the negotiating team is attempting to strike a balance between ensuring stronger protection and getting as many nations to sign.
Influence Rating: 3/5
The OECD AI Principles
2019 will see countries belonging to the Organisation for Economic Co-operation and Development adopt a non-binding set of principles defining some values which should guide AI development. These principles state that AI systems must be transparent, explainable, robust, safe, and secure. They should also have accountability mechanisms. These principles state that AI systems should also contribute to economic development.
ProsThese principles have been used to shape AI policy initiatives all over the world. For example, the EU AI Act will adopt the OECD legal definition of AI. The OECD monitors and tracks national AI legislation and conducts research on AI's impact. The OECD has a global network of AI experts who are actively involved in research and best practice sharing.
Cons The OECD as an international organisation has a mandate to promote economic growth and not regulate. This is what Smuha says. Phil Dawson, the head of policy for the AI platform Armilla, said that translating high-level principles into practical policies is a huge task.
Influence Rating: 4/5
The Global Partnership on AI
In 2020, the Global Partnership on AI was created by French President Emmanuel Macron and Canadian Prime Minister Justin Trudeau. It is an international organization that can share information and research on AI. It also fosters international research collaboration and helps inform AI policy around the globe. The organization is made up of 29 countries. Some are in Africa, South America and Asia.
The value of GPAI lies in its ability to promote international research and collaboration, according to Smuha.
Some AI specialists have asked for an international organization similar to UN's Intergovernmental Panel on Climate Change, to share knowledge and conduct research on AI. GPAI could be the answer. After launching with much fanfare, the organization has kept a low-profile and hasn't released any publications in 2023.
Influence Rating: 1/5
The EU AI Act
The European Union has finalized the AI Act. This is a comprehensive regulation that regulates the "high-risk usages" of AI systems. The bill, first proposed in 2021 would regulate AI systems in sectors like health care and education.
Pros The bill can hold bad actors responsible and prevent the worst overreaches of harmful AI. It will do this by issuing large fines and prohibiting the sale and usage of non-compliant AI technology in Europe. The EU is the first to introduce a comprehensive AI regulation. The EU regime is likely to become the de facto AI standard in the world, as companies from non-EU nations that wish to trade with the powerful trading bloc must adjust their practices in order to comply.
Cons Many of the elements in the bill are controversial. The EU will be subjected to intense lobbying by tech companies. It will be at least two years before the bill snakes through the EU's legislative system.
Influence Rating: 5/5
Technical Industry Standards
Dawson says that technical standards set by standard-setting bodies are going to play a more important role in translating regulation into simple rules for companies. Once the EU's AI Act is passed, for example, companies that meet certain standards will automatically comply with the law. There are already many AI standards, and there will be more in the future. ISO has developed standards for risk management, impact assessments, and managing the development of AI.
Pros These Standards help companies translate complex regulations into practical measures. Dawson says that as more countries write their own laws on AI, the standards will be a great help to companies in building products that are compatible across different jurisdictions.
ConsMost of the standards are generic and can be applied across industries. Companies will need to translate the standards to suit their industry. Dawson says that this could be a huge burden for small business. The question of whether engineers and technical experts should draft rules on ethical risks is a bone of contention. Many people are concerned that policymakers will just pass on difficult questions regarding best practices to the industry standards development, says Dawson.
Influence Rating: 4/5
The United Nations
The United Nations (which counts 193 member countries) wants to be an international organization which can support and facilitate global AI coordination. To achieve this, the UN appointed a new technology ambassador in 2021. In the same year, UNESCO, a UN agency, and its member countries adopted a voluntary AI ethical framework. The member countries committed to, among other things, introducing ethical impact assessments of AI and assessing the environmental impact.
Pros Only the UN has the power to influence AI policies in countries from the Global South. The UNESCO AI Ethics Framework has had a huge impact on developing countries that are still learning about AI ethics. China and Russia have signed the principles, despite being largely excluded from Western AI ethical debates.
Cons This raises questions about how serious countries are when it comes to following the voluntary ethics guidelines. Many countries, such as China and Russia have used AI for surveillance. When it comes to technology, the UN has also had a mixed record. The first attempt by the UN to coordinate global tech was a disaster: the diplomat selected as technology envoy, was suspended just five days after a scandal involving harassment. The UN has been trying to develop rules for lethal drones, also known as killer robots, for years.
Influence Rating: 2/5
————————————————————————————————————————————————————————————
By: Melissa Heikkilä
Title: Our quick guide to the 6 ways we can regulate AI
Sourced From: www.technologyreview.com/2023/05/22/1073482/our-quick-guide-to-the-6-ways-we-can-regulate-ai/
Published Date: Mon, 22 May 2023 17:39:00 +0000
Leave a Reply