Abstract 

As students navigating the complicated world of artificial intelligence, I imagine that many of my fellow Laidlaw researchers have observed debates about how AI threatens the future of work and has the capacity to replace many jobs previously performed by humans, for better or for worse. However, this narrative obfuscates the reality that there are already immense amounts of human labor embedded in AI. Online crowdsourcing platforms are just one forum where humans are paid to complete many of the computational tasks that power AI softwares, and these crowdsourcing jobs are often categorized as contract work, thereby freeing them from complying with traditional labor laws that stipulate minimum wages and that limit overtime work. 

Often, our discussions about the jobs that will be lost because of advanced forms of technology ignore the fact that AI itself is extremely labor intensive. Cade Metz writes for the New York Times: “A.I. is learning from humans. Lots and lots of humans.” People—particularly low-wage workers, both in the United States and abroad—have provided the basis for the AI revolution, training technological systems by annotating digital images, identifying figures and objects in satellite photos, and conducting the repetitive work that provides the basis for ChatGPT and self-driving cars. This work can have detrimental effects on the people providing these services, particularly those whose tasks involve medical videos, violent images, and graphic content. 

Research Objectives & Questions 

My research objectives include first obtaining a holistic knowledge of the current state of the AI industry and the labor market and attempting to decipher the intersections between the two, including the types of jobs that AI currently has created or threatens to replace. Through my research, I seek to contribute a more nuanced perspective to the reductive narrative that currently exists around AI and the future of work by investigating the thesis that AI will replace human jobs and replicate human knowledge. I also hope to answer the following questions, along with many others that I imagine will emerge throughout the course of this summer:

• Has the subdivision of technological tasks, the outsourcing of labor to crowdsourcing platforms, and the rise of AI softwares contributed to a return or a renewal of industrial labor exploitation?

• Will the AI revolution and new machinery streamline the processes of work and eliminate menial tasks performed by humans, or will it instead create a demand for additional jobs to address technological difficulties? 

• Tracing the idea that Ellen Ullman characterizes as the “original sin” of AI—or our belief that the mind can act like a computer, and the computer can act like the mind—is there ever truly a way that AI can replace human labor if it is incapable of conducting unconscious and subconscious processes? 

Background 

The release of ChatGPT in 2022 and the subsequent explosion of AI softwares has sparked debate about the types of work that machines can perform and how human labor will be altered as AI becomes a larger part of our everyday lives. As Kate Crawford articulates in Atlas of AI, “Rather than representing a radical shift from established forms of work, the encroachment of AI into the workplace should properly be understood as a return to older practices of industrial labor exploitation that were well established in the 1890s and the early twentieth century” (58). Crawford’s book has played a huge role in informing my research thus far and in drawing connections between modern and historic forms of labor exploitation. Following her precedent, I hope to use the study of AI systems to understand and subvert the limiting argument that new technology will create a more just labor industry through its supposed elimination of menial tasks.

There are many crowdsourcing platforms that are working to fuel the seemingly all-consuming desire to advance AI systems, including Amazon’s Mechanical Turk (MTurk), as well as CrowdFlowder, SamaSource, iMerit, Figure Eight, Microworkers, Clickworker, and many others. I will identify the key hazards of crowdsourcing work, along with baseline statistics on minimum wage, relations between workers and requestors, and desires for additional forms of support for workers on each of the unique platforms. Companies like iMerit rely on humans to identify pornography and graphic violence in various forms of media, and oftentimes, those workers are separated into distinct rooms to protect other workers from exposure. Further, there is a large international aspect of AI, as a large percentage of low-paying jobs on crowdsourcing platforms are outsourced to laborers overseas. Crowdsourcing labor plays a unique role for workers in the Global South that is wholly unique from workers in the Global North, and therefore, there must be an acknowledgement of neocolonialism and the impacts of globalization when discussing the disparate impacts of AI on different populations of workers. 

Because much of the literature on AI is quite recent, I plan to use many modern resources to trace the history of AI systems and their development over time, including Augmented Exploitation, What Computers Can’t Do, Humans Need Not Apply, Atlas of AI, and a myriad of other resources. 

Methodology 

After conducting a literature review and gaining a broad understanding of the modern state of AI and crowdsourcing work, I will investigate the labor practices and compensation policies of different crowdsourcing platforms to identify key issues and hazards that contribute to labor exploitation. Centering my research in globalization theory and in prominent philosophical and economic theory, I will look at trends in crowdsourcing data to determine the utilization of technological work and low-wage workers across various industries. Engaging with AI will be a critical part of my research, and I plan to ask generative AI systems a series of questions to gauge if and how human bias is embedded into the product. I will also survey the messaging and advertising used by different AI companies to assess the degree to which they obscure the role of workers in their platforms. 

An additional method that I hope to employ is to incorporate firsthand experiences to my research through the Turkopticon, a communal online space for workers behind AI to access resources, aid, and support. The platform is designated for workers for MTurk, a crowdsourcing website that completes computational tasks for AI and classifies its workers as contractors, permitting Amazon to pay mere cents for tasks that workers complete and avoid compensating its workers adequately to meet overtime or minimum wage requirements. Workers use the Turkopticon to share personal testimonies about working for MTurk, and in order to preserve their anonymity, the website obscures their identities. Although the administrators of the website are clear that they do not want research that exploits or undermines the experiences of the site users, they specify that they are open to research inquiries, so I plan to email the site to determine the possibility of gathering personal testimonies from their users, provided their information is kept wholly anonymous and included in my final research product based on their discretion alone. 

Potential Impact 

Through my research work, I hope to contribute to the existing scholarship on AI by centering the workers who are frequently ignored or forgotten when discussing mechanization and technological development. I hope to connect modern forms of labor exploitation to historic trends and apply decades and centuries-old methods for countering worker abuse to our current situation, updating these methods as necessary, in order to propose some preliminary solutions for protecting workers in this rapidly evolving technological revolution.  

Resources & Support Needed 

Because I hope to connect modern forms of labor exploitation emerging with the AI revolution to historic forms of labor exploitation, I will be leaning heavily on my mentor, Professor Joseph McCartin, to find resources that chronicle the recent history of worker abuse and the legislation that has been enacted to counter this mistreatment, both in and outside of the United States. Using his expertise on North American labor history, I hope to identify gaps in labor laws that fail to address the new challenges that AI poses to worker safety and well-being. 

Additionally, I would absolutely love to hear any personal stories with AI. A central element of my research project is attempting to disrupt the typical narratives around AI that dominate large news sources, and the way that I hope to do that is by focusing on personal testimonies, including the ones from other Laidlaw scholars. If you have any experiences encountering bias in AI systems or any testimonies on how human labor is embedded in AI, I would absolutely love to hear from you. And finally, if you have any suggestions or questions on my project, please reach out!