AI Regulation – Like Platting Fog!

The race is well and truley on to regulate AI, which stands for Artificial Intelligence for anyone who has not been online for the last few years, and in that glorified nomenclature lies the regulators biggest challenge. AI refers to a broad field of computer science that involves the development of intelligent systems capable of performing tasks that typically require human intelligence.

There are scarce few politicians who will be able to hold a light to the intellect required to navigate these waters lucidly. Which leaves them even more than normal at the mercy of lobbyists. Or as I am sure a few will, resort to asking a large language model, such as ChatGPT (a large language model derivative of AI) for guidance, just about sums up what we can expect in terms of first generation regulation.

It will be crucial to get some ground rules in place or a lot of time and lost innovation momentum is going to be incurred. Developing effective AI regulations requires deep technical understanding, which will be lacking among policymakers and regulators. Bridging the gap between policymakers and technical experts is a crucial first step to formulating informed and effective regulations.

Parking for now whether AI is actually intelligent or even conscious as some suggest (and that of the policy makers!), the fact is the genie is out of the bottle and no world of regulation is going to get it back in. While humanity divines what this new ephemeral ‘Utility’ actually means for our future health, wellbeing and security, regulation will both help and hinder.

The Challenges

AI poses several challenges due to its unique characteristics and rapid development in a world that dos not relate to data consistently from one continent to the next. For DATA is the raw material of life for AI.

Take for example China with its social scoring and promiscuous data sharing and lack of digital privacy compared with that of Europe and the US. This in itself is likely to have already given China a competitive edge to train and develop large language model algorithms across enriched data sets that are not permitted by other countries. This equates in visceral terms to one country permitting human experimentation where all others abide by the World Medical Association’s Declaration of Helsinki. This declaration sets forth ethical principles for medical research involving human subjects, emphasizing informed consent, protection of vulnerable populations, and the importance of balancing risks and benefits.

Perhaps not a bad place to start if we are talking about a significant dimension of humanity, the data of its personages.

Stepping back a bit, the complexity and diversity of AI alone presents a veritable swarm of moving targets For AI encompasses a wide range of technologies, algorithms, and applications that will be hard to fit in any meaningful way to a comprehensive regulatory framework. AI systems can be highly complex, often involving deep neural networks or machine learning models with millions of parameters. This illustrates how difficult it will be to fully understand and predict behaviour outcomes.

Technologies accelerating speed of advancement means that this complexity and diversity will evolve at a similarly dizzying pace, outpacing the development of any regulatory frameworks. New AI techniques, algorithms, and applications are emerging Dailey. How will regulators keep up with the latest advancements and potential risks associated with them?

We already struggle with technological advancements lack of Standardisation, so how can we possibly expect to achieve an AI standard(s) on the equivalent of this digital quicksand? There is no unified standard or consensus on how AI should be developed, deployed, or regulated. Different countries and organizations have varying approaches, priorities, and ethical considerations when it comes to AI. This lack of standardization complicates the establishment of global regulations.

As mentioned above in the data privacy example, the ethical and bias Concerns that have already been exhibited resulting in unintended discriminatory behaviour. This is giving AI the benefit of the bout for now, as they learn from historical data that may contain inherent biases. Regulating AI to ensure fairness, accountability, and transparency while avoiding unethical use is another lens the regulators will need to consider which adds complexity because determining what constitutes ethical AI practices will be subjective and culturally dependent.

They will also have to balancing innovation and regulation. AI has the potential to bring about significant societal and economic benefits. Excessive or overly restrictive regulations may impede innovation and hinder the adoption of AI technologies, potentially slowing down progress in areas such as healthcare, transportation, and education.

It’s not as if the regulators are dealing solely with the subject matter itself, but the Global economy with its cross-border challenges. AI operates globally, and regulations need to consider cross-border data flows and international collaboration. Coordinating regulatory efforts between different jurisdictions and reconciling varying legal frameworks adds complexity to the regulation of AI.

While these challenges make it difficult to regulate AI, it is important to address the risks associated with AI technology through a combination of ethical guidelines, industry self-regulation, and targeted regulatory interventions that balance innovation and societal well-being.

The UK Position

The UK government has expressed its intention to develop AI regulations and has released a first of its kind roadmap to an effective AI assurance ecosystem. While the specific regulations are yet to be finalized, there are some potential issues and concerns including:

• Scope and Definitions – AI is a broad field with various subfields and applications, making it difficult to create regulations that cover all aspects without being overly broad or restrictive. Striking the right balance in defining the scope and terminology is crucial for effective regulation.

• Ethical and Bias Considerations – Critics argue that the UK’s regulatory plans should place more emphasis on preventing discriminatory outcomes, ensuring fairness, and mitigating biases that may arise in AI algorithms. They suggest that the regulations should provide clearer guidelines on ethical considerations and incorporate diverse perspectives.

• Enforcement and Compliance – Questions have been raised about the enforcement mechanisms, potential penalties for non-compliance, and the capacity of regulatory bodies to oversee and monitor AI systems effectively. Adequate resources and expertise will be required to enforce the regulations successfully.

• Innovation and Competitiveness – Striking a balance between regulation and fostering innovation is a key challenge. Overly restrictive regulations could stifle innovation and impede the development and adoption of AI technologies. There is a need to ensure that regulations do not unduly hinder technological advancements and that the regulatory framework promotes innovation and competitiveness.

• International Harmonization – As AI operates globally, harmonizing regulations across jurisdictions is crucial. Ensuring consistency and compatibility with international standards and regulations can facilitate cross-border collaborations and the responsible use of AI technologies. Coordination with other countries and organizations is necessary to avoid fragmentation and regulatory conflicts.

In Summary

Yes, it is possible to regulate AI. As AI technologies continue to advance and play increasingly significant roles in various aspects of society, there is growing recognition of the need to establish regulations to address the potential risks and ethical concerns associated with AI.

While regulating AI poses challenges, there are ongoing efforts by governments, organizations, and experts to develop frameworks and guidelines for responsible AI development and deployment.

For now, pragmatically, regulations can focus on different aspects of AI, such as data privacy, algorithmic transparency, accountability, fairness, safety, and ethical considerations. They can also address specific applications of AI, such as autonomous vehicles, facial recognition, or healthcare systems. Regulating AI involves setting legal requirements, standards, and guidelines that govern the development, deployment, and use of AI technologies.

This is a meal of many courses that will need to satisfy many pallets that cannot be rushed. AI regulation will need the involvement and input from many stakeholders, including policymakers, technologists, ethicists, legal experts, and the public. Engaging in multi-disciplinary discussions and taking into account diverse perspectives will be essential to creating effective and comprehensive regulations that balance the benefits of AI with potential risks and societal impact.

If you found this interesting, perhaps read my earlier article on AI – The Risk of Cultural Sterilisation by Artificial Intelligence

UPDATE 18th July’23

The evolving regulatory landscape for AI in the European Union (EU) and the United Kingdom (UK), highlights different approaches and potential implications for organisations, particularly in the financial services sector.
The EU has approved the text of its AI Act and plans to implement it within the next year and a half, eventually leading to subordinate legislation. By 2025, technical standards related to AI will also be established, although this process takes longer due to the complex nature of developing such standards in such a bureaucratic structure.
The UK, on the other hand, adopts a pro-innovation stance towards AI regulation. Although it is still consulting on the establishment of technical standards, it aims to become a world leader in AI. Currently, the UK is fourth on the global AI index, making it a significant player in the AI field. Despite recent advancements in the US, the UK holds a strong position and continues to strive for leadership in this area.
Policy changes around AI are frequent in both the EU and the UK but differ in their focus. Notably, the principle of clear rationale, which refers to the transparency and understandability of AI decision-making, is emphasized in the UK but not in the EU. This principle could be particularly advantageous for the UK’s financial services sector, where the UK has a traditionally strong presence, as transparency in AI-driven decisions is often highly valued in this sector.
This reinforces the importance for CIOs and organisations to stay updated with the evolving regulatory environment, particularly in fast-paced fields such as AI. Both the EU and UK’s approaches to AI regulation present opportunities and challenges that organisations need to understand and navigate.