In his 2018 book, AI Superpowers, venture investor Kai-Fu Lee predicted the world would evolve into Cold War-style digital power blocs, one led by the US and the other by China.

The two economic giants would achieve overwhelming dominance in developing artificial intelligence models, Lee argued, because companies in those countries have more venture funding, more scientists, and, above all, far more data than those in other nations.

Since AI tends toward monopoly — “better products lead to more users, those users lead to more data, and that data leads to even better products and thus more users and more data,” Lee wrote — the US and China would leap out to “massive leads,” reducing other nations to digital client states.

That prophecy hasn’t daunted tiny Singapore. When it comes to AI, the Southeast Asian city-state — home to 5.6 million people crowded onto a landmass about a quarter the size of Rhode Island — punches well above its weight. In London-based Tortoise Media’s Global AI Index, which assesses AI capability in 62 countries across more than 100 different metrics, Singapore ranked third behind only the US and China.

The island nation is leveraging its mammoth container port and bustling airport to offset a dearth of domestic data. Its giant banks and scrappy “super app” companies like Grab and Sea are using AI and data analytics to drive regional and global growth strategies. 
 

In many ways, Singapore offers a case study in how small and medium-size nations can keep pace in an escalating AI arms race. Its experience suggests small states might even have advantages over large ones in the Big Data era — and can be uniquely agile in unlocking AI’s power.

“If you’re a small country, you need to be smart, you need to be faster, you need to be nimble,” says Oliver Tonby, senior partner in McKinsey’s Singapore office. “Singapore shows what a proactive and dynamic government can do.”

Singapore was one of the first countries to adopt a national AI strategy, in 2019. Last December, Deputy Prime Minister Lawrence Wong (who’s since become prime minister) updated and expanded those initiatives in a policy framework billed as “National AI Strategy 2.0.” As part of that framework, the government has allocated US$743 million over the next five years to boost the country’s AI capabilities. 

Josephine Teo, minister for digital development and information, says that Singapore’s aims are modest. “We aren’t trying to be an AI superpower,” she insists. “We don’t need to be.” Instead the city-state hopes to position itself as a kind of digital Switzerland, trusted by players in both power blocs. 

To an extent, Singapore has achieved that goal already. Its top tech startups are bankrolled by venture investors from both the US and China. Singapore’s data centres host cloud infrastructure for Amazon, Google, Microsoft, and Meta Platforms — but also China Telecom and Alibaba.

In Singapore’s central business district, Chinese tech giants like TikTok parent ByteDance and online retailer Shein jockey for office space alongside Google, Amazon, and IBM.

The question, far from resolved today, is whether any of Singapore’s nascent AI players will build global footprints of their own. 
 

A hallmark of many of Singapore’s AI initiatives is an effort to leverage its status as a global hub for travel, finance, and cargo, with the hope of developing and monetising AI specific to those industries.

Singapore’s Changi Airport is one of the world’s busiest; more than 59 million travellers passed through last year. Changi uses AI to screen and sort baggage, and to power facial recognition technology for immigration clearance.

Today, that technology makes Changi one of the world’s most efficient airports: In the future, it could conceivably power algorithms that Singapore could market elsewhere. AI plays an equally vital role at Singapore’s sprawling container port, the world’s second-busiest after Shanghai. Last year, the port handled a record 39 million TEUs (or 20-foot equivalent units, each one roughly equal to a standard shipping container).

The Port of Singapore uses AI to direct vessel traffic, map anchorage patterns, coordinate just-in-time cargo delivery, process registry documents, and more.

Those capabilities helped operators cope with a sharp surge in demand earlier this year as global shipping lines rerouted away from the Red Sea to avoid Houthi attacks, according to David Foo, assistant chief executive at the Maritime and Port Authority of Singapore. And the port’s operators have become so adept at using AI that they are preparing to license their management systems to other shipping hubs.

Singapore’s giant banks, too, have embraced AI with gusto. DBS Bank, Southeast Asia’s largest lender, boasts a team of nearly a thousand data scientists, analysts, and engineers, up from only 25 in 2017, according to DBS chief analytics officer Sameer Gupta, who says bank executives were inspired by the way Formula 1 racing teams use data. At F1 races, “AI strategy is now equally, if not more important than your car,” says Gupta.

Similarly, at DBS, data staff have gone from the back office to “the front lines.”Singapore also hopes to offset its small size through increased collaboration with the rest of Southeast Asia, home to more than 680 million people. One obstacle to that aspiration: the region’s extraordinary linguistic diversity.

Southeast Asia spans 11 countries whose residents speak more than 1,200 different languages. Southeast Asians complain that LLMs — large language models — created in Silicon Valley don’t work well in their languages because, for all their speed and power, those models are mostly trained in English.

Singapore sees that as an opportunity. In December, the Singapore National Research Foundation said it would allocate US$52 million to develop a ChatGPT-like AI that would be the first ever tailored to Southeast Asia’s languages and cultures. The model, dubbed SEA-LION (short for Southeast Asian Languages in One Network) is an open-source engine designed to translate 11 major languages.

That seemingly impossible task has been assigned to a team of 25 researchers working out of a small office at the National University of Singapore under the auspices of AI Singapore, a government-sponsored agency created to promote AI adoption by Singapore’s private industry. 

It’s a tiny team operating on a pittance (for comparison, OpenAI has raised US$14 billion to date). But Darius Liu, a SEA-LION team leader, argues that bigger isn’t always better for translation engines — and that Southeast Asians can’t rely on the goodwill of American tech bros. “What if the Western powers decide to turn their models off?” he asks. “If your languages are Thai, Malay, Tamil, or Tagalog, you’re going to get left out.” 
 

Even before the AI boom, Singapore was one of the world’s biggest digital hubs. The country is hyper-connected to the rest of the world via 25 undersea cables, with plans to add 14 more over the next decade.

But if Singapore is to thrive as a global hub for data, the city-state must build more data centres — no small challenge for an island nation where land is scarce, energy is expensive, and it’s sweltering all year round. 

Right now, Singapore hosts more than 70 data centres and 1.4 gigawatts of capacity. In 2019, the government declared a moratorium on new centres, citing concerns about the space and power they consume. The result: Operators rushed to build data centres in Malaysia and Indonesia, which were more than happy to accept their investment dollars. 
 

This May, Singapore issued a new road map allowing for as much as 530 megawatts of new capacity. But to win state approval, the new data centres will have to meet much stricter sustainability and green energy standards. 
 

Tim Rosenfield, co-founder of Sustainable Metal Cloud (SMC), a Singapore-based cloud services provider, sees the city-state’s resource constraints as the kinds of limits that spur innovation. Rosenfield and colleagues at an Australian engineering company called Firmus developed a technology that uses liquid immersion to cool GPUs, instead of the energy-intensive air-cooling used in traditional facilities.

Firmus builds immersion tanks dubbed HyperCubes that can be installed in a standard cargo container and shipped anywhere in the world.

In June 2023, Firmus teamed with ST Telemedia Global Data Centers, a giant Singapore data centre operator, to found SMC, which is retrofitting HyperCubes in ST Telemedia facilities in Singapore and the rest of Asia. The HyperCubes can reduce data-centre energy use and carbon emissions by up to 50 per cent, according to a leading benchmark of computing power consumption.

Simon Chesterman, vice provost at the National University of Singapore and a senior director at AI Singapore, argues that Singapore’s modest size encourages collaboration between the public and private sectors. Such collaboration is particularly vital in a nation whose government has yet to enact formal AI regulations.

Chesterman says Singapore is searching for a governance model somewhere between the approaches of the US, which eschews regulation in favour of market-led growth; the European Union, which has prioritised data privacy; and China, which emphasises social stability and state control.

For Singapore, Chesterman says, the challenge is to avoid under-regulation, putting citizens at risk and undermining public trust, and overregulation, which could scare away foreign partners and stifle innovation. The city-state may never be an AI superpower. But it might help convince the rest of the world that in the age of Big Data, small states too can think big.
 

Source: This article was first published on Fortune.com. © 2024 Fortune Media IP Limited. All Rights Reserved. Reproduction in whole or in part without written permission is prohibited.

Photo: © Getty Images