TACOMA, Washington — The great promise of AI is that it gets us to solutions to core problems faster than has ever been possible in human history. This is good because a major problem just got worse: The COVID-19 pandemic was predicted to push an additional 88 million to 115 million people into extreme poverty in 2020, with the total increasing to as many as 150 million people by 2021. Poverty now represents the number one priority for the United Nations Sustainable Development Goals for 2030. Narrowly applied AI could directly tackle some of the underlying challenges.
Supercharging Efforts Using AI
A strategy for reversing the economic trend toward inequality might be found in one of the most exciting developments in global development policy: the embrace of randomized experiments to determine how small changes make a significant difference, an approach pioneered by Economic Nobel laureates, Abhijit Banerjee and Esther Duflo. What if the world supercharged that effort with AI?
AI can improve the way we collect and measure the data from these trials and lower the costs of goods and services while empowering the community that requires help. These three vectors of information gathering, cost reduction and community empowerment represent a meaningful foundation for how the AI community can contribute to global development challenges. There are currently several case studies that show the promise of such an approach.
Better Responding to Need
In regard to information gathering, AI can improve the measurement of randomized trials. For example, aerial drones can monitor the health of vegetation at scale through a combination of near-infrared imagery and computer vision algorithms. This would allow farmers to more effectively test novel crop treatments and identify the best ones. Researchers at Stanford have also developed a way to assess global poverty through computer vision by analyzing the light patterns of satellite imagery, using algorithms to detect areas without access to electricity. Signals like these allow aid organizations to better identify and respond to need.
In regard to healthcare, inexpensive cameras powered by AI can provide timely diagnoses of illnesses without the need for doctors to be present. Even housing data can be better tracked through AI. The World Bank is currently using computer vision to assess the vulnerability of shelters in disaster-prone regions and provide structural interventions to prevent collapse. These are just a few examples of how AI can generate crucial metrics needed to better assess interventions that alleviate the structural issues of poverty.
Lowering the Cost of Essential Needs
When it comes to lowering the costs of essential needs, Charles Kenny, the British development economist, once wrote that “the biggest success in development has not been making people richer but, rather, has been making the things that really matter—things like health and education—cheaper and more widely available.” The promise of AI means this may be achieved,
The billions spent in developing autonomous vehicles can also be used to help lower the cost of delivering the most important necessities to the global impoverished. After all, distribution is the main obstacle to food security, as the world does in fact produce enough food to feed its population. Further, automation in factories can help lower the cost of producing essential goods through the advancement of robotics and the algorithmic optimization of supply chains.
Additionally, education can be greatly enhanced at low cost by leveraging the mechanism we apply toward personalized ad recommendations to develop personalized educational syllabi, something that was found to be far more effective than enforcing a single rigid syllabus, through the same randomized control trials performed by Banerjee and Duflo. Almost every aspect of the production of necessary goods and services can be improved at scale with AI.
Finally, in order for these interventions to have lasting structural changes, the communities served must be empowered to sustain the interventions. The good news is that the underlying structure for community development already exists. Many companies based in developing countries currently annotate and label datasets needed to train AI algorithms. Empowering these workers to develop additional skills higher up on the AI pipeline can give them considerably more control over its implementation and provide valuable skills for a future economy transformed by these technologies. The academic community can also contribute to this development by promoting the creation of datasets that can be directly applied to the problems that the global impoverished face.
The poverty-stricken have been most vulnerable to climate change, war and disease, among other catastrophic events. Developed countries have benefited the most from the use of technology and AI. It is time to use the promise of AI to aid and empower the impoverished communities that need it the most.
– Daeil Kim