AI is Being Trained Using ‘Digital Slaves’ in Developing Nations Because of Course It Is

An image of an AI an to illustrate AI digital slaves and digital slavery

Artificial intelligence is learning thanks to a “vast, hidden industry” of “digital slaves” from developing nations. This is according to a report in the November/December edition of Wired UK which details how millions of workers from the Phillipines to Colombia are being paid “pennies” to put the ‘intelligence’ in AI.

Australian company Appen is named as one of the key players in this global industry. The company is a data services provider based in WA that outsources data labelling work to some of the world’s cheapest labour markets. Their clients include Amazon, Facebook, Google, and Microsoft to whom they provide the services of 1 million people globally.

It works like this; AI algorithms are trained on large data sets. In order to ‘teach’ algorithms rules for interpreting and responding to inputs, they first need to comprehensively understand these datasets as categories of things. For example, when you’re asked to ‘prove you’re human’ in one of those captcha tests, you’re really labelling things to train algorithms to better identify, say, traffic lights. It’s dull, tedious work, but someone has to do it if AI is going to function.

The people best suited to do that work are those who can be bought very cheaply. Journalist Niamh Rowe details how a vast community of Appen workers reside in Venezuela. This is a country that has been in economic freefall since 2017 and where inflation is 800%. Over six million people have fled in the past six years and are living as refugees in nearby nations.

“In Venezuela, you don’t live, you survive,” said one Appen worker, Oskarina Fuentes.

The work is remote and pays between 2.2 US cents and 50 US cents per task. An hour and a half of work might net workers like Fuentes USD $1. Appen tasks are posted randomly and you need to be online to snap them up or you risk not having work for the day, something she notes is increasingly an issue.

“Fuentes works on a laptop from her bed, glued to her computer for over 18 hours a day to get the first pick of tasks that could arrive at any time. Given Appen’s international clients, days begin when the tasks come out, which can mean 2 am starts,” Rowe writes.

It’s not just Appen who trains AI in this way. Other companies, like Clickworker and Scale AI also employ people to do similar tasks in places like East Africa, India, and even refugee camps in the Middle East. OpenAI has also been accused of running similar low-paid operations in Kenya.

“One needs to work five or six hours to complete what effectively amounts to an hour of real-time work, all to earn $2,” one worker in Pakistan told Rowe.

“In my point of view, it is digital slavery.”

The data labelling industry is currently valued at USD $2.22 billion and is expected to grow to $17.1 billion by 2030, according to consulting firm Grand View Research.

Over the past few years, experts have been cautioning that AI is now at the forefront of a new world of ‘digital colonialism’. The term was coined by researchers at Stanford University in 2021 and explored in the book ‘The Costs of Connection’. In 2022, the MIT Technology Review ran a series of investigations into ‘AI colonialism’ which detailed the experience of Appen workers in Venezuela.

“European colonialism, they say, was characterized by the violent capture of land, extraction of resources, and exploitation of people—for example, through slavery—for the economic enrichment of the conquering country,” journalist Karen Hao has written.

“While it would diminish the depth of past traumas to say the AI industry is repeating this violence today, it is now using other, more insidious means to enrich the wealthy and powerful at the great expense of the poor”.

Appen has been cagey in reports over how exactly it is dealing with the issue of so-called digital slavery. Their website states that they are “harnessing AI for good to create a better world.” Perhaps tellingly, the company recently published a blog post in which they state that they have recently signed the United Nations Global Compact, part of which relates to stamping out slavery.

“We’ve established a global ethical sourcing and modern slavery policy, which we’ll strictly enforce within our own business and among our suppliers,” the blog reads. It explicitly states that it has “no tolerance” for slavery in its labour force.

In June, the European Union and the United States announced progress on a common “code of conduct” in the AI sector which would be voluntary for the industry to uphold. UNESCO has also previously warned that governments should pay attention to the potential for exploitation in the AI sector given the vast wealth discrepancies between nations.

Fuentes herself, who has now become something of an icon of the ‘digital slavery’ problem, has previously spoken out against characterisations of herself as such.

“I don’t feel like a slave to either Appen or AI,” she told El Pais earlier this year.

“We are slaves to the Latin American system.”

She explained that it is the system of global inequality that allows people like her to be exploited, not specifically AI or technology.

Related: All Students Will Be Taught How to Use AI Next Year (By Humans)

Related: Australia’s Secret Weapon in the Battle Against Bushfires This Summer Could Be AI

Read more stories from The Latch and subscribe to our email newsletter.