It’s a big world. Somebody’s gotta label it – AEI – American Enterprise Institute: Freedom, Opportunity, Enterprise31st May 2019
Forbes ran a fascinating story yesterday about off-shore data labeling, the process by which images are “tagged” for analysis by computers. Data, oceans of it, are the primary raw material for machine learning (smart machines getting smarter), and the tech titans can’t get enough of it. The best way of acquiring accurately labeled data (e.g., pictures of automobiles that show up as “sedan” versus “SUV” in the database) is to get human beings sitting in front of computer screens to do the labeling manually. The alternative is to develop programs that analyze data independently based on broad parameters set by human programmers. This second method, for the time being, is both technically more difficult and yields a less accurate result. Score one for the human beings.
Given the infinite variety and quantity of data that requires labeling and the ravenous appetite that all kinds of commercial entities have for that data, the big high-tech firms are turning to developing countries for educated but inexpensive labor. Across Asia and Africa, firms like Google, Microsoft, Salesforce, and Yahoo! are employing low-cost tech workers to do the work of labeling the world. These jobs typically pay between $3 and $6 per day which, as the author says, is “chump change” to Big Tech but often the best job on offer for workers in developing countries.
This is a trend worth paying attention to. Richard Baldwin, an economics professor at the Graduate Institute (Geneva), former Bush administration staffer on the Council of Economic Advisors, and author of “Globotics Upheaval: Globalization, Robotics and the Future of Work,” believes white-collar jobs are at risk of being drained away, much like manufacturing jobs were before them, but faster. There’s no need to build a factory or move equipment — all it takes is a laptop and Wi-Fi connection. At the other end of the labor-product pipeline, platforms like Upwork are helping to identify and aggregate highly skilled workers in developing countries, making them available on a situational basis to businesses and entrepreneurs everywhere. It’s the gig economy, international version.
Baldwin believes rapidly spreading high-speed internet and significant recent improvements in online translation programs (e.g., Google Translate, DeepL) are making it possible for well-educated foreign workers, even those with limited English, to take on routine, computer-based work like accounting, computer programming, and engineering at a fraction of the cost of domestic workers. The chart below shows the relative pay scales for several white-collar occupations in the US, UK, Poland, and China.
The potential “value-capture” in the pay differentials between developed and developing world professionals is significant. Forget illegal immigration or H1-B visa workers: “telemigration” means that, in principle, substantial portions of the service sector on which the American middle class depends for “good jobs at good wages” may be exposed to foreign competition and subject to off-shoring.
The history of automation has been overwhelmingly positive for working conditions and living standards, with new wealth generating new jobs in a virtuous cycle. But it is worth remembering that the transition from farm to factory took a century to complete and the one from the factory to the office took a generation, allowing time for the middle class to build literacy levels or otherwise retool for new occupations. Even in those time horizons, the changes were wrenching for communities and families. Baldwin makes a compelling case that the globotics revolution is something new. It holds the potential for immense wealth generation, but it will likely stress our economy, workers, and politics in ways and at a speed that we’ve never seen before.