Assessing Oxford Insights’ Government AI Readiness Index in 2023

And its relevance to Philippines' AI readiness

An attempt to measure the AI Landscape

After it was shared by our chairman, Doc Ligot, in the Responsible AAI council, I delved into the Oxford Insights’ Government AI Readiness Index for 2023 report. A preliminary review of the report piqued my interest due to its comprehensive overview of the current global state of AI, despite its primary focus on assessing the AI capabilities of individual countries' governments. However, certain sections prompted me to contemplate whether the report might also provide insights into other evolving facets of the AI landscape, which I am continually striving to understand. This exploration is directly related to several projects I am planning for 2024, as I gear up to contribute to AI policy-related work and educational seminars, with a particular focus on the Philippines, my home country.

Looking at the Philippines’ global standings makes me wonder how they studied it. Let’s start looking at the methodology they used.

Pillars, Dimensions and Indicators

Diverse sources were employed to gather data on AI dimensions, facilitating cross-regional comparisons. Each dimension encompasses 'indicators' through which the authors assessed and determined the ratings, forming the foundation for the rankings of each country. To assess their method, I compiled a table outlining the dimensions and indicators, incorporating descriptions and sources referenced in the report.

The 'pillar-dimension-indicator' method appears to offer a solid framework for evaluating countries' readiness for government AI, presenting a conceptually comprehensive approach. However, due to time constraints, I have not yet reviewed the datasets, spreadsheets, and web links that were provided, nor have I examined how the scores were calculated. I've shared a link to the spreadsheet I created for anyone interested in exploring the report's assessment independently. At present, I concur with the authors' efforts in assessing governments' AI adoption readiness. While the report is informative, its completeness is debatable, especially considering the ongoing challenges in AI safety engineering (a.k.a the AI alignment problem). The report's utility and accuracy could be enhanced as advancements in this domain become available and be measurable for each country.

Some high-level takeaways from the report that directly affects the Philippines

I organized this section to prioritize the three pillars most crucial for the Philippines to focus on. This material is not exclusively intended for governmental use; I believe these insights can be valuable to anyone. Even as an ordinary individual seeking to understand the direction of local AI developments, you can use this information to adapt and refine your approach. Personally, my goal is to advocate for solutions that address these issues, with the hope of creating a significant impact in any possible form.

I believe that the final section of the Data and Infrastructure Pillar is spot-on.

“Amid the generative AI boom, which puts the spotlight on major risks like privacy, labour displacement, and misinformation, it’s important not to underestimate the effects of the existing global digital gap. While the emergence of generative AI models holds the potential for significant improvements in public services for countries in the lower income bracket, the associated risks must be acknowledged. Without a solid base of data and infrastructure, countries may find it challenging to develop domestic generative AI capabilities, potentially leading to reliance on foreign technology. This reliance could introduce additional hurdles, including the unavailability of AI tools in local languages and the potential for biases in AI models. Addressing these challenges becomes essential for fostering equitable and inclusive advancements in AI readiness.”

Having indirectly faced challenges related to data and infrastructure during my AI safety research, I've reached conclusions similar to those in the first section, but from a different angle. The necessity of being in a conducive research environment became evident to me, particularly given the Philippines' issues like poor internet connectivity, unstable cybersecurity, and limited support for research and development. This personal experience has led me to believe that both data and infrastructure are vital in any country's endeavor to effectively utilize AI technology.

Data is often compared to the new oil. No country can effectively compete without high-quality data that accurately reflects their values, aspirations, and is representative of their population. The significance of this point might not be immediately apparent to those new to AI, but the disparity will become clear in terms of who can produce the necessary datasets for their intended applications or solutions.

Moreover, local infrastructure has a profound impact on data management. Countries with better access to data centers will likely surpass those lacking such facilities. Take the Philippines as an example; due to data transfer requests predominantly traveling underwater, the country suffers from slower data transfer speeds, which in turn hampers productivity. I contend that these disadvantages, accumulating over time, lead to significant setbacks.

Quoting the conclusion of the Technology Sector Pillar:

“Even considering these outperforming middle-income countries, the large disparity in tech sector readiness is concerning. If a country’s domestic tech sector is too immature or lacking in human capital or innovation capacity to create adequate AI tools, governments may be forced to turn to foreign companies, likely in higher-income countries, to procure AI services. This both stunts the growth of the domestic tech sector and can have even more dire consequences for AI-enabled public services, which may be improperly trained on foreign data not representative or relevant to a country’s context.”

Countries with underdeveloped technology sectors are at a significant disadvantage, almost on the verge of a crisis, particularly when compared to nations that have made effective investments in human capital. Such investments have led to the emergence of companies and industries that produce advanced products and services, contributing to high-tech economies. In the case of the Philippines, finding a clear solution to this challenge is complex. A practical approach might involve adopting the Pareto principle—focusing on engaging a sufficient number of curious and capable individuals who are willing to confront the anticipated numerous obstacles. This approach could, hopefully, spark a transformative movement.

However, the feasibility of this idea in practice remains uncertain. Even if the Philippines overcomes data and infrastructure issues, there might not be enough bold intellectuals to drive a revolution in innovation. My assessment could be incorrect, of course. Perhaps the key lies in an effective educational campaign to bridge these gaps. Consequently, I am also exploring ideas for initiatives that could support this effort.

On the Government Pillar and its relevance to the Philippine Government

“The number of AI strategies released per year is trending downward, but the picture looks more diverse.”

The global decline in the release of governmental AI strategies is not a major concern for me at present, considering the many uncertainties involved. My research into the AI alignment problem underscores a significant issue: the lack of adequate solutions, similar to the absence of necessary safety engineering for controlling AI systems. While I lack concrete data to support this, I suspect it is indirectly linked to the absence of a consensus on AI governance implementation. Notably, major labs are beginning to integrate governance into their agendas, a move largely influenced by the increasing focus on AI safety.

This situation presents both opportunities and risks. The Philippine government should deliberate on appropriate regulatory measures. Ideally, these should address the country's current technological, data, and infrastructure challenges, encompassing the support of relevant strategies, laws, and organizations.

Should the Philippines struggle to develop the necessary technology, data, and infrastructure sectors, the government must, at a minimum, safeguard the nation, its citizens, and its interests, while not impeding innovation. Strengthening the research and development sector would be a crucial step in this direction.

Closing thoughts

In conclusion, my assessment of Oxford Insights’ Government AI Readiness Index for 2023 brings me to expand upon a statement from its introductory page:

“Governments are not only working to regulate AI and foster AI innovation, but also striving to integrate this technology into public services.” [Quoted from the report]

This perspective aligns with my previous discussion on the emphasis on data and infrastructure pillars. I want to underscore the significance of controlling such revolutionary technologies. I don't foresee a slowdown in AI progress globally over the next three years. This period is crucial: it provides an opportunity either to capitalize on the benefits of this technology or to fall behind in the race entirely.

Should the Philippines—or any country—participate in this race? Absolutely, but with a focus on responsible implementation. In fact, I should consider writing a blog post about how Responsible AI implementations could be effectively applied in real-world scenarios.

While I find the report valuable, as I've noted, the challenge of AI safety engineering remains a significant concern. Despite my desire to find reassurance in such reports, I must acknowledge considerable uncertainty in the pillar-dimension-indicator method due to the lack of a comprehensive theoretical framework for safe AI implementation. Was my time spent analyzing this report beneficial? Largely yes, as it provided a deeper insight into global developments and prompted me to consider how adapting to AI challenges will vary—being straightforward for some countries but immensely difficult for most.

---

Thank you for reading. Please consider subscribing to receive notifications of future posts; I plan to publish at least two AI-focused blogs each week. Let me know thoughts in comments section!