Why “AI Slop” Hides the Debate About AI’s Role in Society

Summary

In a commentary published in npj Artificial Intelligence, Daniel J. Singer argues that the phrase “AI slop” often conceals rather than resolves important debates about artificial intelligence. He identifies two common uses: an optimistic “quarantine” view, which treats bad outputs as exceptions that should be separated from useful AI applications, and a critical “contagion” view, which treats derivative or careless outputs as evidence of what AI fundamentally is. The term appears to judge a particular output, but it can also smuggle in a broader judgment about the technology without making that judgment explicit. Singer notes that low-quality work existed before AI, while AI changes the scale, speed, cost, profitability, and detectability of synthetic content, creating distinct concerns including possible effects on future training data. He separates the question of whether an output is good from the harder question of whether it should be produced by a machine, such as in care, education, courts, or hiring. The article does not call for banning the term, but asks people and institutions to state whether they are criticizing an output or taking a position on AI’s social role. Using arXiv’s response to unchecked AI-generated submissions as an example, Singer argues that a quality-control rule can be interpreted as a broader stance on AI. He concludes that these underlying choices about accountability, human involvement, and shared reality should be debated directly.