“Middle powers” can thrive in the age of AI, says Eric Schmidt

AN URGENT QUESTION for national policymakers as the artificial-intelligence revolution takes shape is what to do if their countries do not have cutting-edge “frontier” AI models run by companies similar to OpenAI and Mistral.

AI is increasingly dividing the global order into the haves (countries with the resources and skills to create their own frontier-level models to help accelerate national priorities) and the have-nots (all the others). As America and China set the standard in a tight AI arms race, how can the rest of the world harness the technology’s promise?

Most countries will not have their own national AI champions. Instead policymakers will have choices to make. Models that are small, bespoke, computationally lightweight, open-source and designed for specific use cases? Or large, general, computationally intensive models that may not be fully aligned with their nation’s context? Western or Chinese models? Brick-and-mortar data centres or providers in the cloud? Sovereignty and identity will both play a role in answering these questions, for trusting in, say, the security of governmental AI running on a foreign-sourced model in a remote data centre. And countries will rightly want their AI to have a strong cultural identity and language diversity.

There is no dodging these questions. No matter how small your state, this technology promises to improve economies, ways of governance and social fabrics. All told the effects will be enormous. Goldman Sachs predicts AI will boost labour productivity growth globally by around 1.5 percentage points, representing almost $7trn in extra global GDP over ten years.

In “AI Playbook for Small States”, a report published in September by the Forum of Small States, it is clear that greater adoption will happen in the public sector alongside key industry sectors in countries that embrace innovative applications of AI. The Mauritius Cane Industry Authority, for example, uses AI in agriculture to support precision farming and water management, while Rwanda’s government is working with the World Economic Forum to use AI in health care.

I would like to offer my own playbook for a category of countries that have the capacity to do something meaningful with AI—that is, those with disproportionate capital or large enough domestic labour and consumer markets, plentiful talent and a high demand for AI. These are often called “multi-aligned middle powers” or “geopolitical swing states”.

These countries should first find a niche somewhere along the AI value chain, from regulation to software to data centres. Britain, for example, is positioning itself as a leader in AI governance. Its AI Safety Institute gets ten times the funding of its American counterpart. Saudi Arabia launched a National Semiconductor Hub in June, focusing on simpler chips than those of market leaders. Ireland is making use of its clean-power abundance by building large data centres and shipping out processed data as a product.

Second, swing states will need to use their local context for comparative advantage. How should each country differentiate itself? In what industries does it already have an edge? Leaders should make use of demographics, foster trust and double down on distinctive domain expertise. Populous countries such as India and Indonesia, which have enormous aggregate data pools, can exploit government-held datasets to spur innovation that might uniquely benefit their populations while pushing the wider AI enterprise forward. However, to tap such gains, governments must do the often very hard work of cleaning up their public datasets to make them AI-ready, including instituting clear data-governance policies and data standards that work across jurisdictions.

Viewed in that transnational context, countries with robust regulatory frameworks in data protection stand to benefit more from AI’s spread. Singapore, for example, is positioning itself as a “digital Switzerland”, hosting cloud infrastructure not just for Amazon and Microsoft but also for Alibaba and China Telecom. Anticipated demand led it to increase its data-centre power allocation by 35%. And states can band together in coalitions to build collective advantages and shared frameworks; this is true also for existing groups such as NATO and ASEAN.

The third prescription is about the AI resource that cannot be built or coded: people. Countries will need to lure talent with a compelling global mission and strategic immigration policies. The United Arab Emirates, for instance, is bidding to attract developers by making Falcon, its Arabic large-language model, open-source and free. And it is issuing 100,000 long-term “golden visas” for coders. This strategy seems to be working. The number of AI workers in the country quadrupled between 2021 and 2023. Other talent magnets include investing in “incubators” that fund and nurture startups, and in allocating national computing infrastructure to academics, luring international researchers to local institutions.

In the decades ahead AI will address humanity’s greatest challenges and opportunities, perhaps even resetting a baseline of human wealth and well-being. Just that possibility itself demands that we pursue it. Everyone should benefit from AI’s transformative gains: a more inclusive and prosperous future. For that, we need every country to participate. ■

Eric Schmidt KBE is former chief executive and chairman of Google. With Henry Kissinger and Craig Mundie he is the author of “Genesis: Artificial Intelligence, Hope and the Human Spirit”.