The use of artificial intelligence (AI) is exploding as businesses, governments, and charities adopt it. Some applications are relatively straightforward. AI will be used to recommend products to Amazon customers to purchase and to generate content for his website for the charity.
But AI can also be used in more serious ways, such as helping doctors diagnose illnesses or helping charities raise money.
Worse still, AI could also be used by governments to use facial recognition to spy on their citizens, or to predict in advance which individuals are likely to commit crimes. .
Many institutions, including the EU and the United Nations, believe that some form or regulation is essential, as AI is not going away, and once a technology is invented, it cannot be ‘uninvented’.
Reasons why regulation is desirable include:
ethical considerations: Ensure that privacy and human rights are respected and AI systems are not built with prejudice or discrimination against specific groups of people.
Responsibility and Responsibility: To clarify where responsibility lies in the event of an accident, injury, discrimination, error, etc. caused by an AI system.
Transparency and explainability: Regulations that promote transparency and explainability help build trust and ensure fairness in areas such as finance, law and medicine.
Safety and risk reduction: Regulation of AI in areas where there is a risk of serious harm to humans, nations and economies is considered by many to be essential to avoid an “AI conquers the world” scenario.
However, despite AI’s existence for many years and its recent explosion of capabilities, there are currently few ways to regulate AI. It also doesn’t look like there are many regulations imposed.
The UK government, for example, has no plans to introduce new legislation to regulate AI, and the US government has taken a similar position. Both countries seem content for now to trying to control the use of AI through existing privacy regulations and other laws.
A notable exception to this is the EU. The EU is preparing an AI law likely to be enacted by the end of 2023. It focuses on regulating identifiable risks associated with the use of AI.
Essentially, we categorize AI risks into four types (unacceptable, high, limited, and negligible) and prescribe a set of actions to be taken for each of these four types.
Unacceptable uses, such as using AI to identify potential offenders before they commit a crime, are prohibited.
High-risk applications, such as the use of AI involving financial data, health data, and other personal data, will be heavily regulated to mitigate risk, while limited-risk applications, such as the use of chatbots, will be heavily regulated. It will be very light. This could include informing users that they are interacting with AI to promote transparency and trust.
The UN has not gone as far as the EU, hoping to publish recommendations on the ethics of artificial intelligence that aim to protect human rights and human dignity and provide an ethical framework within which countries can act.
Despite the EU’s efforts to regulate AI, a consistent global regulatory framework is unlikely to emerge anytime soon. Because there are many challenges that must be overcome before that happens.
Below we look at the main obstacles to broader adoption of AI regulation.
Rapid progress in AI technology: The pace of AI development is far outpacing the ability of lawmakers to formulate regulations, at least for the time being. That means the regulations are likely to be outdated by the time they come into force. Therefore, the regulations introduced should not be too vague to be useful, but should be flexible and adaptable. As the rapid pace of development slows and AI technologies and applications become more mature, further regulation is likely to be introduced.
Lack of government expertise: AI technology is rapidly evolving and improving, making it very difficult for regulators to understand how AI works, how it is applied, and the potential risks it poses. Without sufficient expertise in the AI space, the risk of creating unrealistic regulations is high.
Risk of rigid innovation: Overregulation can stifle innovation and slow the development of AI technologies. It is very difficult for governments to strike a balance between promoting innovation and protecting against risks. Especially when regulations in one country may transfer expertise or employment to another country. Suppressing innovation can also put a country at a disadvantage from a security and defense perspective.
global consistency Different regulatory frameworks and standards can create confusion and hinder collaboration on AI-related issues. Countries may therefore be tempted to do nothing until a global agreement is reached.
unintended consequences: Regulations designed to address specific AI risks can have unintended consequences that stifle beneficial applications or hinder AI research. Anticipating and mitigating these unintended consequences is a major challenge, and it is not clear, for example, whether EU AI legislation will allow the EU to reap all of the potential benefits of AI.
All of this means that AI will be something of the Wild West, at least in the near future. Data privacy and other laws will continue to apply, so that doesn’t mean organizations can do whatever they want with AI. But it may still be a few years before the technology is domesticated and tighter regulations start to emerge on how AI is used.
