By Mary L. Gray, cross-posting from the Social Media Collective blog
My collaborator, Siddharth Suri, and I spent nearly 5 years studying a nascent but rapidly expanding piece of the platform economy that we call on-demand ghost work. Right now, ghost work — millions of people around the world working in concert with programmers moving tasks through an API — fuels artificial intelligence and the automation of the internet. This work requires people to contribute responses, at a moment’s notice, and benefits most from a dispersed, diverse set of responses more than the steady input of one person responding to a single call full-time. We see a moving frontier, between what machines can and can’t solve, what we call the paradox of automation’s last mile. As machines progress, they solve problems that previously only humans could solve. But with each solution a new problem — or opportunity for machine learning — presents itself. Engineers, using on-demand ghost work, put their heads down and dig into advancing the frontier of automation once again. The humans who used to solve these now automated problems are continually displaced, as economists David Autor among others, have noted. New labor markets open up as we think of new problems that need solving. We could say that automation is a hard problem, not because of its technical barriers but because each time engineers nail a wicked problem, from voice recognition to self-driving cars, we see another social need or desire that we want to address through automation. Herein lies the paradox: we keep making progress only to find new problems to tackle. There are as many automation problems as there are perspectives on what constitutes a social need or desire and time-efficient ways to address them.
As anyone in the thick of the race to automate responses to human needs and desires knows, we are several decades away (at least) from conquering the hardest problems in automation. As we strive to solve problems, the process of drawing on human insight and creativity through ghost work will repeat, resulting in the rapid creation and destruction of labor markets for new types of tasks. Thus, these new labor markets are, by design, extremely dynamic. Even more unpredictable: The land of IoT sensors and devices will further expand to-date unimaginable on-demand services and products delivered through the power of human-driven ghost work. For every sensor informing an individual about an action they could take (e.g., close their refrigerator, pick up a waiting child, help a elderly family member in immediate need), ghost work will offer new services to respond to the call, when and wherever we need it.
The problem generated by the paradox of automation’s last mile is that we treat those piecework, outsourced, now ghost worked jobs as temporary and marginal, always secondary to the “real jobs” in our economy. Ghost work and the critical role of workers driving the on-demand economy illustrate that contingent labor is no longer exceptional. Arguably, it never was. It’s just been undervalued or rendered invisible, overshadowed by the mystifying and dazzling machines we build to do what humans can do.
The reality is that innovations in automation and on-demand economies are completely dependent on human labor because of the paradox of automation’s last mile. Right now, the effort to automate relies on ghost work — people making themselves available to programmers and customers issuing requests for help through an API. Even if one believes most work can be automated, let’s consider the (long!) stretch of time (and all the productive possibilities) between this moment and the singularity as a chance to rethink the structure and meaning of employment. We can no longer afford to ignore the people—whether they work 40 hours or 40 minutes a week—undeniably vital to advancing automation or delivering the goods and services that make on-demand economies work. I think that’s a good thing for all of society to accept.