Behind the AI revolution has grown a lower class of workers invisible to most of us: thousands of underpaid people in the US and around the world who carefully parse millions of pieces of data and images to help power powerful AI algorithms. Critics call them "new serfs".
Why it matters: these workers - the people who mark up data so that computers can understand what they're looking at - have begun to attract the interest of social researchers and other experts. The latter say that these "scribers" may at least partly explain the mystery of American income inequality - and perhaps how to solve it.
Context: We think that AI is something omniscient, but this is not entirely true. AI in self-driving cars, such as those based on sensors, can take fantastically detailed pictures of the streets and recognize hazards of all kinds. AI can be fed any driving situation and it can process it. But companies developing self-driving technologies need people to tell them what the AI is looking at: trees, stoplights, or pedestrian crossings.
- Without human markup, AI is stupid and can't tell a spider from a skyscraper.
- But this does not mean that companies pay good money to scribers. In reality, they are paid as the lowest paid workers.
- Companies from the USA claim to pay such workers between $7 and $15 an hour. And, apparently, this is the upper limit of pay: such workers are attracted on crowdsourcing platforms. In Malaysia, for example, the average pay is $2.5 per hour.
Look wider: the winners are AI companies, most of which are in the US, Europe and China. The losers are workers from rich and relatively poor countries who are underpaid.
How companies manage those who do markup: Nathaniel Gates, director of Alegio, a Texas-based crowdsourcing platform, says his firm deliberately keeps work to the simplest, most mundane tasks possible. And while this reduces the chances of workers to improve their skills - and get paid better, Nathaniel Gates argues that at least they are "opening doors that were previously closed to them."
- «We create digital jobsthat didn't exist before. And these places are coming to the guys whom automation has forced out of farms and factories, ”Gates told Axios.
However, some experts say that such practices create disparities in the AI economy.
- In a new book "Ghost Jobs" Mary Gray and Siddharth Suri of Microsoft Research argue that markup workers are a prominent part of the economy's most dynamic industries.
- «Economists have not yet come up with how to evaluate this market,” says Gray Axios. “We valued such work as durable goods (which benefit over time - ed. note), but in reality it is a collective mind - that's where the main value is.”
James Cham, a partner at Bloomberg Beta venture fund, thinks that AI companies are playing the difference between the low wages of “markers” and the huge, long-term profits from the products that result from this work.
- “Companies get long-term benefits, while employees are paid only once. They are paid like serfs, paying only a living wage. And the landlords get all the profits because that's how the system works,” Cham told Axios.
- "This is one big speculation"
What's next: Gray says the market can't increase the wages of data-tagging workers on its own.
- In an era when outdated political and economic rules do not work, and societies wear out, experts need to figure out what the earnings of such workers should be.
- What people get paid it is "a matter of morality, not just economics," concludes Gray.
Deepen:
Translation:
Editing: Alexey Ivanov /
Community:
Source: habr.com