More than two centuries ago, textile workers in Nottinghamshire, England, began destroying machinery. This 1811 Luddite uprising emerged amid economic hardship, the Napoleonic wars and disputes over wages, skilled labor and the quality of goods. This machine-breaking was targeted at employers whose production methods threatened workers’ livelihoods. However, the Luddites also petitioned officials, organized public protests and wrote to industrialists. Fundamentally, they were people who understood the machinery they opposed and had specific objections as to how it was being used and how it was affecting them.
Today, calling someone a Luddite doubles as an insult for someone frightened by or unaware of technology. I find that interpretation particularly frustrating because it entirely misses the question that their struggle continues to raise: Who gets to decide what technological progress means for the people living through it?
That question belongs at the center of our arguments about artificial intelligence.
The variety of AI doomerism I object to takes the possibility of exploitation and turns it into an argument for personal helplessness. In conversations about AI, admitting to using a tool can feel like confessing allegiance to the company behind it or endorsing every use to which the technology might eventually be put.
When I say “AI doomerism,” I mean the reflexive suspicion of AI use. Researchers studying misaligned models, workers and unions defending their livelihoods and people challenging the status quo surveillance system deserve to have their arguments taken seriously.
Criticize the companies. Reject unsatisfactory products. Question what data they collect and who is allowed to use it. However, the leap from those objections to a blanket refusal of AI must also be scrutinized, because your unopened laptop does very little to inconvenience the people making decisions about how these technologies will be adopted.
A useful technology allows someone to accomplish more with a fixed amount of effort. Previous examples of useful technologies that multiplied productivity and allowed for greater specialization include: the cotton gin, the steam engine and the development of electrical transmission lines. These changes might become an employer’s excuse to cut staff or increase expectations, even though history has shown again and again that technology creates jobs and strengthens the economy. More importantly, technological advancements often save time, by teaching workers new skills or by making previously impossible projects possible. Treating exploitation as the only imaginable outcome concedes a remarkable amount of control to the same people these ‘doomers’ disapprove of.
Scientific AI further complicates blanket generalizations about AI. AlphaFold, an AI biology company, predicts a protein’s three-dimensional structure from its amino acid sequence. By 2022 it had already developed a public database of more than 200 million protein sequences, each sequence giving a researcher a unique starting point for investigating how proteins work and where possible cures exist.
At Oxford, researchers combined AlphaFold predictions with laboratory evidence to determine the structure of a protein that shows promise for a malaria vaccine, providing vital information to future disease prevention work.
AlphaFold’s success illustrates just how much the category of “AI” encompasses. A protein-structure predictor and a generator of disposable internet content deserve judgments based on what each actually does; the label of “artificial intelligence” will not do that thinking for us.
The same specificity and knowledgeability can make opposition movements far more effective; rather than blanket hostility towards the technology, workers can instead seek to adjust technological adoption to benefit them.
For example, the Writers Guild of America’s 2023 agreement secured concrete protections grounded in specific and well-reasoned demands. Companies could not require writers on covered projects to use AI and specified that AI-generated material could not be used to undermine their compensation or writing credits. Furthermore, writers retained the option to use AI with the studio’s consent.
Circling back to the Luddites, their grievances force us to ask who controls a technology, who benefits from it and who deals with the inevitable disruption. Learning to use AI does not answer those questions. It does, however, help people recognize useful applications and understand model limits amongst a sea of exaggerated claims. This technical understanding becomes more consequential when paired with collective bargaining, public movements and demands for accountability.
People who distrust AI companies have a unique set of reasons to participate. They can better demand agreements before a workplace adopts a system, they can organize for a share of the productivity it generates and they can argue for limits on use cases that threaten their wellbeing. Participation should include the power to refuse a particular application. But it should also involve understanding what is being refused and working toward an alternative; if you destroy your machines and refuse to participate or find alternatives, you will simply be left behind.
You do not owe an AI company your enthusiasm, and you have no obligation to give them your money or your trust. But if you believe their decisions could reshape your livelihood, you have a stake in influencing those decisions. Skeptics often supply the best questions that enthusiasts have overlooked. Therefore, if you distrust the people steering the ship, you have every reason to fight for a hand on the wheel.
Luke Dolan is a member of the Class of 2027 in the School of Industrial and Labor Relations. He is a staff writer for the Arts & Culture department and can be reached at lpd39@cornell.edu.
