A New Industrial Revolution: Will AI Widen or Close the Income Gap Between Rich and Poor Countries?

One late January night in 1812, a group of men with concealed faces smashed knitting machines with sledgehammers in a textile workshop near Nottingham, England. This raid was one of many carried out by "Luddites"– those who feared that the technological change brought upon by the Industrial Revolution would result in the replacement of workers by machines. We now know that the Industrial Revolution was fundamental to the dramatic increase in wealth the world has experienced since. While virtually all nations have benefited from this surge in wealth, some nations have benefited more than others: global inequality today is nearly twice what it was in the early 1800s.

Two hundred years later, we are potentially facing another such revolution, with rapid developments in artificial intelligence that promise to reshape our economy. As such, there is both fear and optimism about how the future will look when this technology reaches its full potential worldwide.

With AI already impacting certain occupations and industries, one can't help but ask: how will this transformative technology affect closing the gap between rich and poor countries–one of the most important challenges of our time?

The answer: it depends. In particular, it depends on the extent to which AI will result in displacement (which, in the extreme, could result in certain occupations and industries becoming obsolete), or augmentation (i.e., generating productivity enhancements to such an extent that demand for labor increases). In all likelihood, it will be a mix of both, though one hopes that the effects of augmentation will outweigh the effects of displacement. That hope is aligned with a recent report by McKinsey and Company that estimates that generative AI could add trillions of dollars annually to the global economy by boosting productivity.

To investigate this question, we use occupational exposure scores built by Felten et al. (2021), which link different applications of AI (like image recognition or translation) to ratings of human abilities and use this linkage to create a rating of how exposed different jobs are. This data provides a useful metric for jobs within the US, but to tackle the question of cross-country dynamics following the roll-out of AI technologies, we want to be able to compare the overall exposures of different nations. While we don’t have data on the job breakdown of every country, we do have detailed and reliable data on export baskets for each nation, which we can use as a proxy for their industrial composition (ideally, we would want to use standardized production data, but it's typically limited to fewer countries and frequently inaccurate). A limitation of our approach is that we are limited to only tradable goods, and as such, services are excluded. Nevertheless, their exclusion shouldn't change our overall conclusion.

To estimate how exposed nations are, on average, to AI, we use Felten’s industry-level AI exposure scores and create a score for each country based on a weighted average of their 2021 export baskets according to industry shares (with data from Harvard's Growth Lab Atlas of Economic Complexity). We plot these values against the income per capita of countries. The results are presented in the figure below.

We take two important insights from this figure. First, there is a strong positive correlation between AI exposure and income per capita: wealthier countries do, in fact, have higher overall exposures to AI. This is consistent with the repeated findings of researchers that, within the US, the jobs most exposed to AI tend to be higher-income jobs, rather than lower-income ones.

Second, despite the positive correlation, there is important variation across AI exposure for any given level of income per capita. As such, a country like Kazakhstan is much less exposed to AI than China is, though both of them are at roughly the same level of income per capita ($11,492 vs. $12,720 in 2022).

Knowing these two stylized facts, what effect should we expect AI to have on global inequality? There are a number of forces at play here, with competing effects on technology, so we find it useful to consider a few simple and diverging stories for how global inequality may be affected.

First, AI rollout could result in increasing inequality across nations: If AI has a positive effect on productivity, and in turn on growth, then it’s not difficult to imagine that countries that currently have less exposure to AI will miss out on the gains from the technology. Of course, harnessing this growth might depend on access to certain infrastructure (computers, servers, etc.) and the availability of talent. If firms in poor countries are unable to access credit to invest in those (as is likely the case) then economic growth will occur only in a select few rich countries, and the countries without exposure or access to infrastructure will not be able to catch up, resulting in global inequality rising precipitously.

It’s possible to imagine another story, where the rollout of AI will result in reducing global inequality. Suppose a country’s access to relevant infrastructure isn't as important or the investments to acquire that infrastructure do not pose an important barrier to firms in developing countries. Consider, instead, what might happen if AI-based technologies and services are widely proliferated such that anyone with a cheap computer and access to the internet can access a wide range of productivity-enhancing AI tools. In the age of an increasingly globalized workforce, AI could allow workers in low-income countries to overcome the various barriers that currently preclude them from accessing the benefits of global economic progress. For example, imagine that AI tutors and training modules could provide engineers in developing countries with the necessary technical expertise and certifications to compete in advanced manufacturing sectors–like that of the semiconductor industry–that they otherwise wouldn’t have had the requisite training for. With larger cohorts of well-trained workers, developing nations will be better equipped to compete on a global scale and attract prized manufacturers. Consider, similarly, the service sector: if improved real-time translation allows workers to more easily overcome language barriers, workers in developing countries could more easily provide services internationally.

While we don’t know which of these two scenarios is more plausible, policy, and particularly multilateral development banks, could play a role in steering the trajectory towards the second scenario by helping workers in developing countries get as much access as they can to emerging AI tools.

Critically, developing countries do not need to be competitive at the frontier of AI progress to reap the rewards of the new technology. Nations do not need to have their own top AI labs training up models or domestic chip manufacturers. Rather, to access much of the benefits of AI, workers in developing countries simply need reliable access to broadband internet, which will bring with it access to all publicly available AI tools.

This is still a non-trivial hurdle, however. Among people living in low-income countries, only 20.0 percent accessed the internet in the past three months (compared to 90.2 percent  in high income countries), and only 44.81 percent of this population even has access to electricity (99.98 percent, in high income countries). The push to expand developing nations’ access to the internet is nothing new. In some developing countries this is not only a matter of investing in broadband infrastructure, but also investing in expanding electrification and even providing wider access to credit to the private sector to get basic computer equipment. The proliferation of AI urges us to scale up these efforts significantly. Without reliable access to the internet a developing nation’s workforce misses out on the far-reaching potential of AI tools for education, training, and skill-enhancing possibilities.

The new industrial revolution is here, and with it, perhaps a unique opportunity for a global productivity boost that could improve the lives of billions of people worldwide. It is imperative that we set all of us, especially developing countries, up for success.