The dynamic and often unpredictable nature of Artificial Intelligence (AI) development has raised significant concerns about its governance. While regulation is key in managing these risks, standalone solutions like the new legislation AI Act in the EU and AI Bill of Rights in the USA is not enough, this is a global concern.
The introduction of solutions like VMware’s Private AI presents a challenge and an opportunity. The challenge encompasses enforced controls and the lack of a centralized model, especially when anyone can develop an AI model or reuse an existing one with comparatively little capital outlay. The opportunity lies in the shift of focus towards corporate responsibility and decentralized control mechanisms, a must-have to manage AI on a global scale regardless of government legislation. This blog delves into why regulation alone is insufficient and what additional steps are needed to ensure AI evolves beneficially and responsibly and the opportunity that Cloud Providers have to assist customers on a responsible AI journey.
The Unpredictability of AI Development
AI technology is characterized by its rapid and unpredictable evolution. Deep learning, for example, has enabled AI systems to make decisions in ways not even fully understood by their programmers. A great example of this is Microsoft’s experimental chatbot Tay, which quickly learned to use offensive language, proving the unforeseen directions AI can take. These are so surprising that researchers are now trying to predict unpredictability and figure out how emergent abilities have occurred.
The Limitations of Current Centralized Regulations
Current regulations will fail to keep up with AI’s fast-paced advancements, regulation is always a consequence, not a precursor, and nearly all regulation is localized and enforcement problematic. Of particular concern is the lag that regulation inherently has, this allows for unintended consequences, for example, the unregulated deployment of facial recognition technology leading to privacy and bias concerns, as seen in systems like Amazon’s Rekognition.
The Global Nature of AI
AI and software know nothing of borders, and this ultimately complicates centralized national regulations, leading to a patchwork of global standards and universal unenforceability or applicability. Even despite the recent shift in awareness, we have seen in 2023, such as the G7 releasing an international code of conduct for responsible AI and the UK international summit on AI safety, it is no good unless we have global participation from all countries, and with the best will in the world is doubtful to ever happen. However, a positive outcome from this realization is the need for an international framework for responsible AI innovation.
Ethical Considerations and Societal Impact
AI’s ethical dilemmas extend beyond what regulation can address. AI-driven decisions impact societal norms and individual rights, such as in AI-assisted criminal sentencing ‘where AI can enhance sentencing consistency, reducing disparities caused by human biases and subjective decision-making’ (Use-of-AI-in-Criminal-Justice.pdf (rahmanravelli.co.uk)). The ‘disparities’ are however a long way off from being solved, as the COMPAS system in the US showed us, demonstrating racial and gender bias in predicting future criminal behavior
The Role of Corporate Responsibility and VMware Private AI
The shift towards decentralized AI systems, as seen in VMware’s Private AI, underscores the need for enhanced corporate responsibility. VMware’s approach to AI, which emphasizes private, decentralized AI operations, needs a model where control and accountability are more distributed. This decentralization can potentially reduce some risks associated with vulnerability and single points of failure, large-scale data privacy breaches, and lack of innovation. Rather, decentralized AI mitigates risks with no single point of failure, it increases data privacy and security, improves scalability and efficiency, resilience to coordinated attacks, and provides diversity of perspectives, solutions, and innovation.
The role of Cloud Service Providers and Responsible AI
Decentralization using Private AI implies companies must establish robust ethical guidelines and accountability mechanisms (some can be found online to get companies started), no single organization should be exempt from understanding and establishing these guidelines, especially when the risk of not doing so is so great.
Regional Sovereign Cloud VMware Service Providers have a unique opportunity with distributed AI solutions, by delivering local compliance and regulation they can ensure AI solutions adhere to regional data privacy laws and other standards. Many Cloud Service Providers (discounting the hyperscale commercial public clouds) are able to deliver customized solutions to meet the customers’ business needs and this tailoring and support makes them far more relevant to local needs and changing markets, particularly in resources contained markets such as Data Scientists and Machine Learning Operations. More importantly, they do not have global portfolios and are able to adapt to change quickly and effectively, essential with the exponential pace of AI development. Of course, data sovereignty is very relevant in the AI world, with regional data storage, processing, and national operations for support, a personalized and collaborative AI development service with advice and guidance helps foster and align responsible AI practices.
Regulation of AI is essential but alone is an insufficient tool for the reasons given above, in the ongoing governance of AI. The emergence of decentralized AI solutions like VMware’s Private AI highlights the need for a new approach that includes proactive faster regulation, international collaboration on a global scale, ethical considerations, both regionally and internationally, and, critically, corporate responsibility as VMware demonstrates in “Responsible AI” aligned to our EPIC2 Values. As AI continues to reshape our world, developing a comprehensive strategy that safeguards against its risks while maximizing its benefits is more important than ever.
READ NEXT: Ethics in the Age of Generative AI: A Closer Look at the Ethical Principles for VMware’s AI
