An unoccupied office space filled with black rolling chairs at cubicles.

The Invisible Labor Force Powering ChatGPT

Speculation on whether artificial intelligence (AI) will replace human labor often leaves out a piece of the puzzle. Tech enthusiasts argue that turning unappealing tasks over to machines will be a boon to human creativity. Skeptics fear that machines are coming to take away human jobs. Many people are unaware that simply making AI “intelligent” requires an invisible labor force working in the growing gig economy. 

To start, AI is not actually intelligent. Large learning models like ChatGPT and GPT-4 are very good at using statistics to predict the next word in a sentence. They do this by being fed large amounts of data that is labeled. These models replicate human writing as well as they do because they are fed so much text that they can predict what the next thing should say. They make their predictions off of the input that you feed it and the words they generate on their own. 


• Learn about the National Domestic Workers Alliance Gig Worker Advocates efforts to negotiate with employers on behalf of gig workers.

• Support Foxglove’s fund for content moderators who have been let go from Facebook. 

Sign the petition created by the Future of Life Institute to learn more about efforts to decrease the risk of AI.

Read Race After Technology: Abolitionist Tools for the New Jim Code by Ruha Benjamin.

The problem is that the AI first needs a human to assign these labels to train data. So many AI developers have turned to platforms like Amazon’s Mechanical Turk (MTurk) to handle this. Globally, the MTurk marketplace has about 500,000 registered workers who receive requests from nonprofits, academics, and businesses to complete small tasks for a small payout (Pew Research Center). 

Gig economy workers often do not earn fair wages. They also are not employees. They are not entitled to benefits. Most are classified as contractors, meaning they pay their own taxes. Movements to reclassify gig employees as direct employees have gained traction but generally fail. Recently, Prop 22 was reinstated in California, defining app-based workers as independent contractors (Investopedia). Gig work sometimes offers a flexibility that traditional jobs don’t. But gig workers also have none of the legal protections or safeguards traditional employees have (Pew Research Center). 

Kenyan laborers hired by a San Francisco company trained OpenAI for less than $2 per hour. The workers were tasked to view graphic text that was so extreme that many of the workers reported PTSD symptoms. This impact was drastic enough that the company ended its contract with OpenAI eight months early (Time).

American companies have long used cheap global labor to get rich. When most people think of outsourced labor, they think of sweatshops, call centers, and other jobs that can be sent to countries where workers earn lower wages or receive fewer workplace safety protections. And aside from simply setting up shop in other countries, many companies turn to independent contractors domestically to complete work that they can’t (or won’t) hire a new employee to do. 

In the book Ghost Work, Mary L. Gray and Siddharth Suri note that AI will always need human assistance, often on-demand and immediate. A task as simple as dispatching a Lyft to a rideshare app user may need a person to verify that the registered driver is the person operating the car before this seemingly computerized task can occur. 
Technology will always involve human labor. Gray and Suri’s solution is to create equitable and dignified employment for the people who do this work. 

As a society, we need to reimagine how we define work and our responsibility to make sure that nontraditional workers have access to the same benefits as direct employees. This could come through the regulation of ghost work (Brookings). We can start by supporting the efforts of content moderators to unionize, such as the newly formed African Content Moderation Union (Time).


• Tools like ChatGPT seem like magic, but they rely on a largely invisible human workforce that makes them more efficient. 

• The gig work economy also includes task-based ghost workers who contribute to the efficiency of machines. These workers receive none of the benefits that traditional employees get.

• Gig work is largely unregulated globally.

2400 1594 Nia Norris
Start Typing
%d bloggers like this: