The Case for Collective AI Refusal

The author compares the rise of AI to the destructive legacy of the automobile, arguing that LLMs are polluting our information ecosystem and threatening human skills. He advocates for a total refusal of AI tools in personal and professional life to preserve human 'metis' and force a necessary slowdown in development. Ultimately, the piece calls for collective resistance through labor organizing and regulation to mitigate the unpredictable harms of a synthetic future.

Key Points

• AI's impact on society mirrors the automobile's history, where immediate convenience led to long-term systemic damage to infrastructure and social health.
• The proliferation of LLM-generated content is degrading the quality of search, professional services, and human communication through 'slop' and hallucinations.
• Using AI tools erodes 'metis,' the deep theory-building and muscle memory gained from performing tasks by hand.
• Collective action, including labor unions and regulatory demands, is necessary to slow down AI advancement and buy time for societal adaptation.
• Individual refusal to use AI is an ethical choice that preserves cognitive agency and personal conscience.

Sentiment

The community is notably divided but leans toward skepticism of the article's call for collective refusal while being broadly sympathetic to its underlying concerns about AI harms. Most commenters acknowledge the real risks of AI-driven information degradation and deskilling but disagree with abstention as a strategy, preferring engaged regulation and selective adoption. The tone is thoughtful and substantive, with many personal anecdotes and historical parallels, though the discussion frequently drifts into tangential debates about cars and urbanism.

In Agreement

• The automobile analogy is apt — just as car-centric design reshaped cities in individually convenient but collectively harmful ways, AI adoption creates self-reinforcing dependencies that make resistance increasingly difficult over time
• AI-generated 'slop' is already degrading the quality of information online, in healthcare, and in professional services, validating the article's concerns about information pollution
• The concept of 'metis' (practical wisdom gained from hands-on work) is genuinely at risk, as AI tools allow people to skip the learning process that builds deep understanding
• Technology companies have historically prioritized adoption over societal impact, and without collective action or regulation, the same pattern will repeat with AI
• The internet itself has already become a 'suffering engine' that monetizes anxiety and outrage, and AI threatens to accelerate this dynamic
• Historical precedents like the USSR's collapse show that societal disruption from technology concentrates power among those who already hold it, not those with practical skills

Opposed

• Collective refusal is impractical because genuinely useful technologies inevitably get adopted regardless of resistance — the utility of transformative technology is undeniable
• Current LLMs are deeply limited and overhyped; comparing them to an existential threat is premature when they cannot perform basic reasoning tasks reliably
• Technologists should stay engaged with AI to influence its development from a position of credible authority rather than ceding the field through abstention
• Intelligence and critical thinking have always been valuable and will remain so — AI makes human judgment and oversight more important, not less
• The article's framing conflates the harms of unconstrained corporate deployment with the technology itself; regulation and selective adoption are more productive than blanket refusal
• Evidence that AI is better at knowledge work without human oversight is very limited; humans with AI agents will be more productive, but the models regress to the mean on strategic insights
• Historical automation fears have consistently overestimated displacement — we keep inventing new types of work, and AI is unlikely to be different