Some of the most crucial work in advancing China's technology targets takes place at a former cement factory in the center of the country's heartland, from the aspiring Silicon Valleys of both Beijing and Shenzhen. An idled cement mixer still stands in the center of the courtyard. Boxes of all melamine dinnerware are piled in a warehouse next door.
Inside, Hou Xiameng conducts a company which aids artificial intelligence make sense of the world. Two dozen young individuals go through videos and photos, labeling just about what they see. That's a car. That is a traffic light. That's bread, that is milk, that's chocolate. That's what it seems like when a person walks. We used to think the machines are geniuses, but now we are aware of the fact that we are the reason for them being genius.
In China, a new generation of low-wage employees is assembling the foundations of the future. Start-ups in smaller, more economical cities have sprung up to apply tags to China's huge trove of pictures and reporting footage. If China is the Saudi Arabia of information, as one specialist states, these firms would be the refineries, turning raw data to the gas which could induce China's A.I. aspirations.
Traditional wisdom says that China and the United States are still competing for A.I. supremacy and that China has certain benefits. The Chinese government widely supports A.I. companies, financially and politically. Chinese start-ups composed one-third of the worldwide computer vision market in 2017, surpassing the United States. Chinese academic papers are cited more often in research papers. In an integral policy announcement this past year, the China government stated that it expected the nation to become the world pioneer in artificial intelligence (AI) by 2030.
Above all, the Chinese authorities and companies enjoy access to plethora of data, as a result of weak privacy laws and authorities. Beyond what Facebook, Google, and Amazon have gathered, Chinese online providers can get more since people there so greatly utilize their mobile phones to shop, pay off for meals and purchase movie tickets.
Still, a lot of these promises are iffy. Chinese newspapers and patents may be guessed. Government money may go to waste. It’s not obvious the A.I. race is really a zero-sum game, in which the winner receives the spoils. Data is useless unless somebody could parse and catalog it.
But the capability to tag that data may be China's true A.I. strength, the only real one which the United States might not have the capacity to match. In China, this brand new sector offers a glimpse of a future that the government has promised: an economic market built on engineering as opposed to manufacturing.
While A.I. engines are super fast students and great at handling complex calculations, they insufficient cognitive abilities that the normal 5-year-old possesses. Small children know that a furry brown cocker spaniel and a black Great Dane are equally dogs. They could tell a Ford pickup from a Volkswagen Beetle, and they know both are cars.
A.I. needs to be taught. It must digest vast amounts of tagged videos and photos until it realizes that a black cat and a white cat are just cats. This is where the data factories and their workers arrive in.
Taggers helped innovation, a Beijing-based A.I. company, fix its automated cashier system for a Chinese bakery chain. Users can put their pastry beneath a scanner and cover it with no help from a human. But almost one-fifth of the moment, the machine had trouble telling wedges from doughnuts or pork buns thanks to saving lighting and human motion, which made graphics more complicated. Working together with photographs from the shop's inside, the taggers got the precision around 99 percent.
Each of the artificial intelligence is developed on human labor. Innovation has fewer than 30 taggers, but a spike in labeling start-ups has made it easy to farm out the job. After, Mr. Liang needed to get about 20,000 photos in a supermarket labeled in three days. Colleagues made it done with the support of data factories for only a couple thousand dollars.
The information factories have been popping up in areas far from the biggest cities, often in distant areas where both labor and office area are inexpensive. Lots of the data factory employees are the sorts of people who worked on assembly lines and building sites in these large cities. But work is drying out, wage growth has slowed along with most Chinese people would rather live closer to home.
Mr. Xi, 36, was out of a project and want to find different ventures moving with basic school classmates when somebody said A.I. tagging. After online searches, he decided it was not super technical however needed cheap labor, something Henan gets in abundance.
Back in March, Mr. Li and his friends set up Ruijin Technology, which rents offices that the size of two professional basketball courts within a commercial park for $21,000 a year. It was formerly the park's Communist Party committee's event space, or so the ceiling lights have been coated with red hammers and sickles.
Ruijin, which means smart golden, today employs 300 workers but intends to expand to 1,000 after the Chinese New Year holiday, when many migrant workers come home.
Unlike workers and company all over the world, Mr. Xi is not worried that A.I. will take his job.
"The machines are not smart enough to teach themselves, " he said.
Hiring is a bigger worry.
Ruijin's pay of $400 to $500 per month is greater than typical at Jiaxian. Some potential job applicants fear that they don't understand anything related to A.I. Others get the work boring.
Jin Weixiang, 19, said he would quit Ruijin following the Chinese New Year and go to market furniture at a physical store in southern town Guangzhou.
However, for a few former migrant workers, the occupation is better than working on assembly lines.
"This was the exact same work, same movement, day after day," said Yi Zhenzhen, a 28-year-old Rubin worker who formerly worked at an electronic part company. "Today I must use my brain slightly."
Most of the time, clients do not inform these info factories exactly what the endeavor is for. Some are clear. Labeling traffic lighting, road signs and pedestrians are often for autonomous driving. Labeling many types of camellia blossoms could be for search engines.
After Ruijin was given the job of tagging the images of countless human mouths. Mr. Li stated he was not certain what it was too. Perhaps facial recognition?
Roughly 300 miles to the northwest, at the Hebei town of Nangongshi, Hou Xiameng conducts her data factory out of the in-laws' former cement mill. Her very first job out of school was tagging faces for Megvii, the Chinese recognition company with a $2 billion valuation that's most renowned for its technology platform called Face+. For this day, a few facial recognition methods recognize her before they do her friends since, she says, "my face is in the initial database."
But life in Beijing was too tough and expensive. She and her then-fiancé, Zhao Yacheng, decided to move back into their hometown and start a data factory. Ms. Hou's parents could pay for desks and computers. They're also renovating the warehouse next door to hire 80 more people.
Like Mr. Xi, Ms Hou does not spend time thinking about the consequences of her job. Are they leading to a surveillance state and a dystopian future which machines will control human?
"Cameras allow me to feel secure," she said. "We're in charge of the machines now."