The Force Awakens: Artificial Intelligence Is Coming – And It Isn’t Going to Put Poor Countries Out of Work
On a scale of starkly polarizing subjects, the global development sector’s pet topics can rarely rival more mainstream hot button issues – with one exception. Discussion around the likely social impact of Artificial Intelligence (AI) can often be as contentious as the debates over politics, race relations or the relative merits of the various Star Wars films. Whether it’s in drawing-room tête-à-têtes or heated conversations during brown-bag sessions, AI has emerged as a predominant talking point – and everyone seems to have an opinion.
Unfortunately, these viewpoints are often shaped largely by hype and yellow journalism – and the Twitter war-of-words that keeps flaring up between AI optimists and alarmists. Lost in these arguments are the nuances of both AI’s economic impact, and its associated risks.
Ex-Machina: The Myth of AI as a Job-Killer
Take the most commonly cited repercussion: AI’s impact on jobs. For decades, Luddites have been sounding the alarm bells for what Keynes called “technological unemployment.” A spate of recent studies has filled their arsenal with compelling-sounding empirical evidence. Some emphasize that 80 percent of the jobs lost in the U.S. manufacturing industry over the last three decades disappeared because of technological change. Others warn that automation threatens to displace 47 percent of U.S. workers, along with 49 percent in Japan and 35 percent in the U.K. McKinsey predicts that job elimination due to automation and AI will be 1.2 billion globally, and Forrester predicts it will claim 9 percent of jobs in the U.S. alone in 2018.
These numbers paint a dire future. But here’s the thing – these worries are not new, and they haven’t been borne out by history. Even a century ago, Gandhi conjured up a catastrophic picture of people dying of starvation due to man’s “craze for machines.” More recently, ATMs were predicted to replace human bank tellers; indeed, in America, the average number of tellers fell from 20 per branch in 1988 to 13 in 2004. However, bank branches grew by 43 percent in that same period, resulting in a net upsurge in demand for human tellers. This piece of evidence exemplifies what economists call the “lump of labor” fallacy – the misconception that labor demand is fixed, and therefore automating some jobs will cause mass unemployment.
For years, researchers and analysts have pointed out why that prediction is flawed. MIT economist David Autor claims that technological progress has always resulted in the net creation of jobs. He posits that only “routine” tasks are susceptible to substitution by machines, thus causing a “skill-biased technical change.” And philosopher Michael Polanyi has asserted that humans will always have a competitive advantage over machines because non-routine tasks, which need flexibility, judgment and common sense, rely on difficult-to-program tacit knowledge.
That’s not to say technology doesn’t cause labor force disruptions: Autor has explored why middle-skills jobs have diminished while both high-skills and low-skills jobs continue to survive. But to his mind, although there will admittedly be short-term workforce polarization and wage inequality, the long-term impact of automation is only structural, in that it shifts the composition of the labor force without overall displacement. The well-known workologist Michael “Dr. Woody” Woodward agrees that automation just augments skills without destroying jobs. And MIT economists Daron Acemoglu and Pascual Restrepo contend that inequality may increase during transitions, but self-correcting forces tend to limit these increases over the long-run.
Dawn of the Age of Machines?
Again, the naysayers hit back, arguing that the fourth industrial revolution is different. With the advent of machine learning (ML), they say, even inherently human tasks can be automated through the brute force of datasets and computing power. Case-in-point: Though hitherto considered non-routine and non-displaceable, driving and medical diagnosis have now been reasonably automated. And algorithms are rapidly taking on the human roles of financial advising, medical assistance, legal aid, customer service, coaching and counselling. So with the widespread adoption of AI and ML, are we really headed toward a jobless future?
Let’s shift gears for a moment and consider the broader backdrop in which these dire predictions are flowing in. Thought leaders such as Stephen Hawking, Martin Rees and Bill Gates portray a doom-and-gloom picture that seems pulled from science fiction, calling AI an existential threat to humanity. They are terrified that in the race to pass the Turing test, researchers might inadvertently hit “singularity” – a point when artificial intelligence becomes smarter than humans – and neglect to code in morality. Elon Musk thinks AI is “potentially more dangerous than nukes.” Oxford philosopher Nick Bostrom takes this thought experiment to its logical extreme and dreads the day that superintelligence will cause the apocalypse. But as MIT robotics professor Rodney Brooks points out, “…common among these people is that they don’t work in AI themselves. … For those of us who do work in AI, we understand how hard it is to get anything to actually work through product level.”
AI’s Real Impact on Emerging Economies
While some tech leaders worry that the rampant development of AI and ML will erode future jobs, others like IBM are investing to leverage the same technology to find business solutions to poverty. Under their Science for Social Good program, IBM announced 12 projects that address the UN’s Sustainable Development Goals by harnessing the power of data science and AI. While one project aims to tackle the global hunger crisis by deploying AI to optimize food transportation logistics, another attempts to simplify complex messages for less literate adults.
Many commentaries reduce the impact of AI on developing economies to the image of a giant robot crushing the livelihoods of the poor. But is that really the case? A recent report by Great Learning shows that India has twice as many jobs in data science and machine learning as jobseekers – a gap that originates from the supply side of the labour market. This gap will only widen as 75 percent of the registered companies in India have invested in (or are going to invest in) ML and data science. The report also claims that 44 percent of all jobs created in 2017 involve banking and financial services, which is the biggest market for data science professionals. This implies an urgent need to upskill Indian youth in data science and machine learning – an industry where 50,000 jobs are currently lying vacant – instead of worrying about future job destruction.
Shifting focus to another region, take the case of developing economies in the Middle East, which derive exponential revenues from the energy sector. They stand to gain the most from process automation technologies, as one of the use cases of AI would enable oilfield engineers to reduce the opportunity cost of time spent troubleshooting problems related to field equipment. The secret sauce to quick service would be cognitive assistance platforms like Amelia, a Siri equivalent for answering technical queries and reviewing machine manuals. Does this sound like a substitute to a field mechanic’s labour – or an addition to his toolkit?
Let’s look further east towards Africa, the promised land for any development economist. The business potential of AI is huge, whether it’s used to diagnose medical problems with healthcare apps like Babylon in Rwanda, or to offer short-term loans to the poor, with automated credit assessments like M-Shwari in Kenya. Companies like OneFi and Tala can make credit available to small business owners in Africa because AI can produce better predictive models by feeding on consumer data, which used to be difficult to gather on the continent.
Anxiety around job loss is a valid concern. But the development sector should not get too carried away by the hype surrounding it. Governments are already responding to AI’s emergence by instituting policies and regulations – some of which could stifle the technology’s growth and limit its benefits. Instead of adding fuel to the fire around reactionary measures, we should be encouraging a more optimistic approach, working to unite governments and AI enthusiasts behind measures that support ongoing innovation while also focusing on the need to retrain workers impacted by the AI revolution. As responsible citizens, it is our task to stay informed and voice our opinions about our governments’ proposed responses to AI – and to avoid falling prey to fear mongering about a technology that could transform industries and improve the lives of billions.
Brinda Sapra is a Research Associate at the Institute of Financial Management and Research (IFMR). The views expressed in this article are those of the author and do not necessarily represent or reflect the views of IFMR.
Image courtesy of Steve Jurvetson via Flickr.