The New York Times rang in the new year with an article about Stanford researcher Timnit Gebru, whose team “analyzed 50 million images and location data from Google Street View” and crunched out a number of interesting observations that previous image analysis couldn’t manage. As you’d expect, project used some heavy duty AI. But human labor was also critical to their success:
But first, a database curated by humans had to train the A.I. software to understand the images.
The researchers recruited hundreds of people to pick out and classify cars in a sample of millions of pictures. Some of the online contractors did simple tasks like identifying the cars in images. Others were car experts who knew nuances like the subtle difference in the taillights on the 2007 and 2008 Honda Accords.
“Collecting and labeling a large data set is the most painful thing you can do in our field,” said Ms. Gebru, who received her Ph.D. from Stanford in September and now works for Microsoft Research.
But without experiencing that data-wrangling work, she added, “you don’t understand what is impeding progress in A.I. in the real world.”
Once the car-image engine was built, its speed and predictive accuracy was impressive. It successfully classified the cars in the 50 million images in two weeks. That task would take a human expert, spending 10 seconds per image, more than 15 years.
According to researchers Mary Gray and Siddharth Suri, this low-paid human labor is the dirty little secret fueling the AI revolution: the “paradox of automation’s last mile.”
Whether it is Facebook’s trending topics; Amazon’s delivery of Prime orders via Alexa; or the many instant responses of bots we now receive in response to consumer activity or complaint, tasks advertised as AI-driven involve humans, working at computer screens, paid to respond to queries and requests sent to them through application programming interfaces (APIs) of crowdwork systems. The truth is, AI is as “fully-automated” as the Great and Powerful Oz was in that famous scene from the classic film, where Dorothy and friends realize that the great wizard is simply a man manically pulling levers from behind a curtain. This blend of AI and humans, who follow through when the AI falls short, isn’t going away anytime soon. Indeed, the creation of human tasks in the wake of technological advancement has been a part of automation’s history since the invention of the machine lathe.
We call this ever-moving frontier of AI’s development, the paradox of automation’s last mile: as AI makes progress, it also results in the rapid creation and destruction of temporary labor markets for new types of humans-in-the-loop tasks. By 2033, economists predict that tech innovation could convert 30% of today’s full-time occupations into augmented services completed “on demand” through a mix of automation and human labor. In short, AI will eliminate some work as it opens up opportunities for redefining what work humans do best. These AI-assisted augmented services, delivered by people quietly working in concert with bots, are poised to enhance our daily productivity but they also introduce new social challenges.
So it may well be that robots/AI don’t end up creating mass unemployment. But that doesn’t mean we won’t end up in a dystopian future, where many people are trapped in very low wage jobs that help propel AI/robotics to new heights, creating vast amounts of wealth for the 1%.
If we don’t want to end up in a dystopian future, let’s make 2018 the year we stop fighting over what robots/AI will do to jobs and start fighting for a better future for all.