A Journey To The Dark Side Of AI Where Workers Endure “Digital Sweatshops”
The advent of AI has created a humanitarian crisis. Workers labor away in digital sweatshops, which are known for their poor working conditions, exploitation, and psychological distress. These workers, who often work in data labeling, represent a new blue-collar working class.
In places like Africa, South and Southeast Asia, and Latin America, workers toil in some cases for twenty hours a day, sifting through thousands of cases per shift. They perform tedious tasks and endure a psychological toll as they review explicit, violent, or disturbing content–child sexual abuse, bestiality, murder, suicide, and torture–to be labeled for machine learning algorithms.
These freelancers and contract workers labor day and night to meet their financial needs, with assignments often coming in during the middle of the night. They must claim them quickly or they’re out of luck.
They toil away on piece-rate microtasks with low pay and no labor rights. AI workers in developing countries earn about US$2.10 per hour, with more than 50 percent earning less than $1.2 per hour. These rates are not commensurate with their roles within the AI economy.
They also deal with payment delays, rejections of contributions, and unpredictable amounts of work on top of intense demands and impractical targets. The result is worsened economic and social disparities globally.
What’s more, workers do not even know what systems they are training. For instance, one study revealed that data labelers in Kenya worked unknowingly for AI training platform Remotasks, a subsidiary of ScaleAI, a company that provides data to Big Tech. They’re often undertrained, as well.
The industry is characterized today by supply chains that limit a worker’s ability to protest their own exploitation—there are generally no minimum wage laws, overtime, or workplace benefits. Legal gray areas are being exploited by Big Tech companies. They find cheap labor in host countries where copyright and licensing protections are at a minimum.
Big Tech does not source its labor ethically, with compliance lacking on privacy laws and equitable compensation. While traditional crowdsourced data annotation relies on precarious, low-paid labor in the Global South—a new kind of slavery—blockchain provides workers all over the world a way out.
Blockchain can shift the industry towards decentralized, automated systems with fair pay programmed into the protocol level, while also eliminating intermediaries and including the unbanked. This could provide jobs to 2 billion unbanked would-be data labelers and validation tasks via cryptocurrency income.
In a decentralized data production model, individuals receive fair pay for creating, trimming, and annotating text data. Smart contracts and blockchain-based ledgers can manage contributions, ensure fair compensation, and maintain high-quality data standards.
Decentralized Autonomous Organizations (DAOs) serve as the governance structure. Such an organization can manage the compensation of data contributors based on demonstrated expertise and reputation. Blockchains also make it possible to pay for micro-tasks involving data labeling and content moderation.
Blockchain enables verifiable, direct systems that promote equity and foster sustainable work in the AI economy. Token-based reward systems, dynamic pricing models, and voting rights for reputable contributors collectively foster trust and a sense of shared ownership.
Decentralized systems provide transparency and have the potential to attract more diverse and invested contributors. Blockchain offers a prospective tool to transform digital sweatshops into dignified and inclusive workplaces in developing nations.
Blockchain technology can raise the income of people who are working in developing nations on labeling AI datasets.
