The rise of artificial intelligence and other new technologies has raised hopes that they might contribute to reducing poverty. History suggests that they need to be embedded in broader frameworks of institutions, regulations, and training.
Every so often, an ascendant technology captures the public imagination, promising widespread lifestyle improvements and advances in poverty reduction. In the 1950s and 1960s, as newly independent countries launched missions of national development after decolonization, the hot technology items were large dams and steel mills, regarded as “temples of modernization” by then–Indian Prime Minister Jawaharlal Nehru. Analyzing the subsequent experience in his book Watering White Elephants? Danish development specialist Ole Therkildsen recounted a litany of missed opportunities: steel mills that shut down just a few months after starting, silted-up dams, hardly any impact on poverty.
A different technology entered the limelight in the 1970s. Ceremonially inaugurated by President Ferdinand Marcos of the Philippines, the “green revolution” in agriculture was built around new and high-yielding seed varieties developed in the lab, needing assured irrigation and large doses of chemical fertilizers and pesticides. Agricultural ecologist Gordon Conway reported in his 1997 book The Doubly Green Revolution how this technology increased crop production but also contributed to rising inequality and had large unanticipated environmental costs, boding poorly for long-term poverty reduction.
By the start of the new millennium, a different set of technological fixes had come to the forefront. It was predicted that digital techniques, relying on “hole-in-the-wall” computers accessible to children living in slums and remote clinics and classrooms, would resolve long-standing problems of ill-health and illiteracy. Once again, the reality did not measure up to the hype. Exploring the outcomes of digital interventions, including some he developed himself while leading a Microsoft team in India, Kentaro Toyama acknowledged in his 2015 mea culpa Geek Heresy that throwing computers at poor people had rarely improved their situations, and in some cases had made matters worse by replacing an old and imperfect way of doing things with a new and imperfect one.
The history of technology-driven fixes to poverty reduction thus presents a complicated picture. Advances in technology stand behind some of the greatest improvements in living conditions, but not every technology delivers on its promise, not everyone benefits, and unanticipated costs are accrued.
Will this time around be any different? A new set of technologies has come to be all the rage on Wall Street and Dalal Street alike. Can this crop of technologies put an end to world poverty?
Nir Kshetri is justifiably proud of having written the first book-length work on the subject. He discusses the antipoverty potential of today’s cutting-edge technologies, including the four to which he devotes chapter-length treatments—artificial intelligence (AI), blockchain, remote sensing, and the Internet of things—as well as others that have recently arisen, such as 3-D printing, 5G telecommunications, autonomous vehicles, genome editing, and virtual reality. All of them can be legitimately bundled together, according to the author, because they are commonly and heavily dependent on big data and speedy computers. He terms them, collectively, Fourth Revolution (or 4R) technologies.
In Western contexts, where these technologies have been around for comparatively longer, some notable evaluations have shown that they have effects that can be deleterious for poverty reduction. In their 2014 bestseller The Second Machine Age, MIT professors Erik Brynjolfsson and Andrew McAfee demonstrated how these technologies inordinately benefit the privileged and tend to exacerbate inequality “between those that have a job and those that don’t, between highly skilled and less advanced workers, between superstars and the rest of us.”
Two other bestselling books also issued warning signals. Martin Ford’s 2015 Rise of the Robots predicted widespread job destruction due to automation. In Weapons of Math Destruction (2016), Cathy O’Neil turned the focus to “victims of algorithms”: people who, after being misclassified by automated processes, are wrongly denied jobs, deprived of benefits, or put in prison. Usually these are poor people, treated as “collateral damage, deemed unworthy and expendable.”
Kshetri is aware of these stark possibilities, and he is sensitive to how they might play out in the comparatively unexplored terrain of developing countries, in contexts as diverse as India, China, Nigeria, and Bolivia. His timely and informative book comes richly laden with a variety of examples from multiple developing country situations.
I have two quibbles with the presentation. First, although he has written extensively on technology elsewhere, in this book the author provides only the skimpiest descriptions of the new technologies. Three short paragraphs on blockchain, for instance, are not enough to enlighten the non-specialist reader.
Not every technology delivers on its promise.
Second, poverty is defined in a manner that makes it difficult to interpret the author’s conclusions. He lumps together under the category of poverty the combined populations of countries classified by the World Bank as low-income and lower-middle-income, which include India and China but also Congo, Somalia, and Haiti. This adds up to half of humanity, roughly four billion people—hence the second neologism in the title, B4B, or bottom four billion. But B4B is an unwieldy and unviable category. The technology needs of Argentina and Nauru are hardly more similar to each other than are those of, say, Portugal and Argentina. And countries are no longer, if they ever were, uniquely rich or uniformly poor. Many Ivorians and Indians are fabulously rich, not just locally, but by global standards; many more, by far, are desperately poor. The technology preferences of Jean-Louis Billon and Mukesh Ambani are hardly the same as those of slum dwellers.
Embedded Technologies
The strength of this timely book is its sheer number of illustrative examples. It shows the range of what the new technologies are capable of achieving. The author catches the trend almost at its onset, however, when everything is just too raw, too recent, and under-investigated. “Supporting evidence,” he notes, “has not been convincing and conclusive.”
Still, like his illustrious predecessors Nehru and Marcos, Kshetri is starry-eyed about his technologies. He claims, “4R technologies have the potential to eliminate the causes of inequality, poverty, and injustice for the B4B.” In support of this assertion, he presents example after example of what the new technologies have been designed to achieve in diverse situations.
Yet without looking at evidence of what ensues down the road, years after an intervention is first introduced, and without concurrently examining counterexamples that show how things can (and do) go wrong on occasion, it becomes difficult to address what many observers, including the historically oriented, would consider the critical question for understanding the worth of a new technology: Under what conditions do new-technology interventions work best for the welfare of the poorest people? How can the new technologies be combined with effective management and supportive organization to achieve the greater good for the largest number?
When Kshetri was working on this book, it was too early to be asking these questions. Most, if not all, developing-country applications had started no more than a year or two earlier. By the laws governing such enterprises, many of these moonshots would fail; such are the expected costs of a risky venture. But it wasn’t clear at that time which of these initiatives would succeed and which would be aborted or mutate. This information is available now and needs to be compiled and investigated.
The author would do yeoman’s service by providing, perhaps in an accompanying web-based portal, updated information about each of these initiatives and the reasons associated with its subsequent growth, demise, or mutation. (As a quick check, I searched the Internet for updates on applications reported in Chapter 2, on AI. The first of these applications, involving AI-enabled smart vision spectacles developed by Aravind Eye Hospitals in India in collaboration with Boston-based nonprofit Vision Aid, is no longer featured on the website of either institution. Harambee, the South African nonprofit, no longer uses AI for job placements for young people. PlantVillage, cited by the author on the basis of media articles published in 2018 and 2019, still serves, according to its website, the same small number of communities in the same handful of locations.)
Developmental outcomes are not about the technology per se, as we should have learned by now; it’s about how a technology is embedded within particular situations. Rather than portraying new technologies as an unmitigated good—a quixotic hope and an unrealistic standard—it is wiser to look for better-performing technology-plus configurations: technology plus institutions, technology plus regulation, technology plus trained and motivated people.
The need for informed action is dire. When claims are made about a technology absent context, absent people, they ring hollow. Meanwhile, negative experiences are accumulating. Laws related to privacy, surveillance, and improper profiling, newly developed in Western contexts, are almost entirely missing in many developing countries. This absence of safeguards enables “ethics dumping,” such as when unsuspecting Bangladeshis and Ethiopians are recruited cheaply and without informed consent for training facial recognition software—which corporations then turn around and sell, often half-baked, to repressive governments that have few misgivings about victims of algorithms.
What is coming, Kshetri promises, “is even bigger” in terms of both downside risks and upside potential. The governments of developing countries need to fashion effective technology strategies for protecting their enterprises and their populations. But more guidance is needed on what a national technology strategy should look like, and more learning needs to be acquired about how to use the new technologies most beneficially. Fourth Revolution makes a good beginning by showing the wide range of possible benefits. As applications of new technologies advance, some more successfully and others less so, these varied experiences need to be evaluated with an eye to gaining traction on the critical question of appropriate technology-plus configurations.
University of California Press