AI giving rise to new type of “ghost work”

Apr 01, 2019, 7:56 AM EDT
(Source: Melvin A/flickr)
(Source: Melvin A/flickr)

The allure of doing away with monotonous, banal jobs with artificial intelligence (AI)-powered automation comes with a spoiler – the technology breathes on what it promises to eradicate. AI algorithms need to be trained on massive datasets, requiring manual labeling, which demands hours of slog in the form of “ghost work.”

The emergence of data factories across smaller, cheaper cities in China, where armies of human labelers toil hard for several hours a day, tagging images and footage to train deep-learning algorithms, is a case in point, writes The New York Times.

As the owner of one of the data-labeling factories in Jiaxian points out, human labelers are “the construction workers in the digital world.” AI, despite its genius at learning and crunching numbers, lacks the cognitive abilities of even a five-year-old and, hence, needs the services of humans to bridge that gulf with the real world.

This kind of invisible work is rightly called “ghost work,” as it remains shrouded beneath the glamor of automation, without contributing anything toward empowering those who drive it and, in fact, creating a “new working class,” reports MIT Technology Review.