Some of Them are Paid Less Than $20 for a 10-Hour Shift

Newsreel Asia Insight #280
July 13, 2024

The global AI market, projected to grow from $200 billion in 2023 to nearly $2 trillion by 2030, is drawing millions of workers into its rapidly expanding sector. This surge in artificial intelligence affects industries like logistics, manufacturing, and healthcare but masks a harsh reality: the widespread exploitation of labour that powers these advanced systems, according to the authors of an upcoming book.

Contrary to what tech giants claim, AI development is not just about autonomous, self-learning machines. Instead, it heavily relies on a large network of low-paid, overworked human workers, write the authors – Prof. James Muldoon from the University of Essex, Prof. Mark Graham from the Oxford Internet Institute and Callum Cant, a senior lecturer at the University of Essex – in an op-ed for Los Angeles Times.

This misunderstanding started in the 18th century with the Mechanical Turk, which seemed like an automated chess player but actually had a person hidden inside controlling it, point out the authors of Feeding the Machine: The Hidden Human Labor Powering A.I. Today, this illusion continues as companies like Amazon call the use of human labour in AI “artificial artificial intelligence.”

The term “artificial artificial intelligence” refers to the involvement of human labour in processes that are typically expected to be automated by AI technologies. Essentially, it acknowledges that what might appear as completely machine-driven or autonomous functions are, in fact, often dependent on humans performing repetitive, manual tasks that help the machines learn and operate effectively. This term is used to highlight the irony that even in the development and operation of advanced artificial intelligence systems, a significant amount of mundane human input is still necessary.

The reality is that AI operates through the hard work of many workers who do boring, repetitive jobs for low pay.

In countries such as Kenya and Uganda, data annotators work for up to 10 hours a day, making less than $2 per hour, with no chance for career growth, the authors point out.

These workers are essential when AI systems falter. For instance, when Siri misunderstands a command or facial recognition fails, these are the individuals who rectify such errors, refining the algorithms, the authors explain.

Their investigations across Kenya, Uganda, Ireland, Iceland, the U.K. and the U.S. have revealed the broad and diverse workforce powering AI. From content moderators and machine learning engineers to data centre technicians and artists, the work is often monotonous and undervalued with little room for professional growth.

Furthermore, when AI is used in management systems, it tends to simplify and standardise tasks, reducing the skills needed and increasing workloads in the name of efficiency. For instance, in the U.K., warehouse workers are closely watched and controlled by complex AI systems that set the speed and type of their work, turning them into just small parts of a large machine.

However, the idea that AI’s dominance is inevitable does not rule out the chance for a fairer work future, the authors argue.

This issue has been raised before. In their book “Ghost Work,” anthropologist Mary L. Gray and computer scientist Siddharth Suri revealed that companies like Amazon, Google, Microsoft and Uber rely heavily on a large, unseen workforce. These workers, performing “ghost work,” handle tasks like flagging inappropriate content, proofreading and designing engine parts, making the internet appear intelligent. About 8% of Americans have engaged in this “ghost economy,” where workers often receive less than minimum wage, lack health benefits and can be dismissed at any time without reason.

This sector, Gray and Suri say, lacks labour laws to protect workers, drawing a diverse group including overworked young mothers, professionals pushed into early retirement, recent graduates struggling to find traditional jobs and minorities excluded from desired positions. They also explore how ghost work can potentially offer opportunities rather than hardships for those involved.

Muldoon, Graham and Cant recommend strengthening and connecting worker organisations worldwide to enhance collective bargaining power. They advocate the use of civil society and consumer pressure to enforce fair labour standards in companies. They also call for government regulations and international agreements to guarantee minimum working conditions worldwide. Lastly, they say, there’s a need for promoting workplace democracy through initiatives such as worker-managed cooperatives and fair governance structures in companies.