The Spark

“The emancipation of the working class will only be achieved by the working class itself.” — Karl Marx

Google's Sweatshops

Jun 10, 2019

The artificial intelligence software tool called Google Assistant is supposed to answer your every question and request, from how to cook a chicken soup to purchasing airline tickets to learning about pyramids, setting timers, booking hotels, etc. In fact, it takes about 121,000 temps, vendors and contractors (TVCs) to develop its "miraculous" artificial intelligence-powered virtual assistant, according to The New York Times. By comparison, Google has about 102,000 full time employees.

In order to become effectively operational, artificial intelligence software needs to learn or understand how human beings communicate, interact and work, and how they naturally function. To do this, the computer is fed vast amounts of information related to languages, sounds, images, human culture, etc.

Each piece of information needs to be carefully "annotated or labeled," for example as a cat, parrot, tree, etc. Each “parrot” needs to be labeled with different colors by painting a parrot photograph on a computer screen with different colors. There are different parrots, so sufficient numbers of parrots need to be colored. The parrots may appear within foliage or amid many different other animals. Thus, in order for the computer to identify a specific parrot among a large number of animals and foliage, a vast array of information needs to be entered into the computer. In the end, millions of images need to be labeled with names and colors.

All this labeling is done by hand! That is, a worker sits in front of a computer and clicks pixel by pixel on an image, tagging a photograph or entering text word by word, spoon-feeding a computer with knowledge all day until the software tool becomes functional. To enter millions upon millions of information points to a computer, Google uses an army of workers without spending much on them, to have this process be profitable.

Thus, these workers who do the labeling are paid low wages and have poor benefits. Google's US-based contract workers labeling information receive $31,200 a year, if they work full time. By comparison, Google charges its own full time employees $38,808 a year to place an infant in its onsite daycare facilities!

On top of that, these workers are pressured to work unpaid overtime. One former Google employee said: “The 40-hour thing was just not respected. It was made clear to us that we were never to log more than 40 hours, but we were never told not to work more than 40 hours. The work that they assign often takes more than eight hours. Every week you fill out a timesheet. One person one time did submit overtime, and that person was later chastised. No punishment, but definitely told not to work overtime.” Sound familiar?

“It's smoke and mirrors if anything. Artificial intelligence is not that artificial; it's human beings that are doing the work,” said one current Google worker. So, there is nothing miraculous in this 21st century artificial intelligence technology: old fashioned exploited human labor and sweat shop conditions are at work.