For example, the opportunities and potential sources of growth that the United States and China enjoyed in the early stages of their economic development are significantly different from those facing Cambodia or Tanzania in today's world.
Recent research by our staff finds that new technologies risk widening the gap between rich and poor countries as automation drives increased investment in developed countries where it is already well established. This could have a negative impact on jobs in developing countries, as they threaten to displace rather than complement the labor force growth that has traditionally favored developing countries. To prevent this gap from widening, policymakers in developing countries need to take steps to increase productivity and improve worker skills.
Results from the model
Our model considers two countries, one developed and one developing, that produce goods using three production factors: labor, capital, and robots. We interpret “robots” in a broad sense to encompass all the new technologies mentioned above. Our main assumption is that robots will replace workers. The “artificial intelligence revolution” in our framework is the increase in robot productivity.
We find that divergence between developing and developed countries can occur along three different channels: production shares, investment flows, and terms of trade.
Production share: Developed countries have higher total factor productivity, which means higher wages. These higher wages encourage companies in developed countries to use robots more, especially if the robots can easily replace workers. Then, as the robots become more productive, developed countries will benefit more in the long run. The more robots replace workers, the greater this gap becomes.
For developing countries that have been waiting for demographic change and have hoped to reap major benefits, things are likely to get much more difficult.
Investment flow: The rise in robot productivity stimulates a strong demand for investment in robots and traditional capital (which we assume are complementary to robots and labor). This demand is greater in developed countries, where robots are used more intensively (the “share of production” channel mentioned above). As a result, investment is diverted from developing countries to finance the accumulation of this capital and robots in developed countries, resulting in a temporary decline in the GDP of developing countries.
Terms and conditions: Developing countries have more unskilled labor than developed countries, and therefore are more likely to specialize in areas that are more dependent on unskilled labor. Assuming that robots replace unskilled labor and complement skilled labor, a permanent decline in the terms of trade in developing regions may occur after the robot revolution. This is because robots disproportionately replace unskilled labor, lowering their relative wages and lowering the prices of goods that are more unskilled-labor intensive. The fall in the relative prices of staple products may act as a further negative shock, reducing investment incentives and leading to a fall in GDP, not only relative but also absolute.
Robots and wages
Our results depend critically on whether robots actually substitute for workers. While it may be too early to predict the extent to which substitution will occur in the future, we find suggestive evidence that it does. In particular, rising wages coincide with a large increase in the use of robots, which is consistent with the idea that firms substitute robots for workers in response to rising labor costs. 
Implications
The productivity gains of robots will increase the gap between developed and developing countries if robots can easily replace workers. Moreover, while these productivity gains tend to increase incomes, they may also increase income inequality for some groups of workers, at least during the transition period and in the long term, in both developed and developing countries.
There is no silver bullet to avoid the divergence. Given the fast pace of the robot revolution, developing countries need to invest more urgently than ever before in improving total productivity and skill levels to ensure that labor is complemented rather than replaced by robots. Of course, this is easier said than done. In our model, improvements in total factor productivity, which account for many institutional and other fundamental differences between developing and developed countries that are not captured by labor and capital inputs, are particularly beneficial because they lead to greater robot and physical capital accumulation. Such improvements are always beneficial, but the benefits are even stronger in the context of the artificial intelligence revolution.
Our findings also highlight the importance of human capital accumulation to prevent disintegration and point out that growth dynamics may differ among developing countries with different skill levels. The situation is likely to be much tougher for developing countries that were hoping to gain significantly from the long-awaited demographic transition. The growing youth population in developing countries was welcomed by policymakers as a potential great opportunity to benefit from the job transition caused by China’s graduation to middle-income countries. Our findings show that robots may take these jobs. Policymakers need to act to mitigate these risks. Especially in the face of these new technology-driven pressures, a dramatic reorientation to rapidly improve productivity growth and invest in education and skills development would be the key to capitalizing on the long-awaited demographic transition.
