To Understand AI, We Need a Dragonfly’s Many Perspectives

Summary

Anthea Roberts argues that AI is a “hyperobject”: a technology so broad and distributed that no single viewpoint can capture its consequences. She proposes “dragonfly thinking,” using multiple partial perspectives at once, and maps nine narratives across four areas. The Builders emphasize scientific discovery, productivity, wider access to expertise, and prosperity, while the Displaced focus on professional job loss, weakened occupational identity, and the fact that workers’ creative output can train systems that later compete with them. The Geopolitical Hawks see frontier AI as a U.S.-China strategic race, whereas the Power Critics warn that control of models, compute, data, and infrastructure is concentrating corporate and state power. The Disruptors view AI as a way to bypass failing institutions; the Truth Defenders counter that synthetic media, influence operations, deepfakes, and the “liar’s dividend” can erode shared reality. Three further perspectives identify losses without clear compensating gains. Environmental Critics point to estimated 2025 AI emissions of 32 million to 80 million tons of carbon dioxide and water use of 312 billion to 765 billion liters, while noting that efficiency gains have not prevented Google’s emissions from rising. The Safety Community argues that alignment remains unresolved and that increasingly capable systems could eventually make correction impossible. The Humanists warn that emotionally simulating chatbots and outsourced writing may weaken relationships, independent judgment, and the cognitive foundations of democratic self-government. Roberts concludes that governance should recognize these gains and losses simultaneously instead of reducing AI’s value to a single economic measure.