“Artificial intelligence is neither artificial nor intelligent.”
Kate Crawford’s statement “Artificial intelligence is neither artificial nor intelligent” is a deliberate interruption. The phrase “artificial intelligence” can make a system sound weightless, autonomous, and almost natural, as if intelligence simply appeared inside a computer. Crawford’s work traces what that label hides: mined minerals, electricity, water, supply chains, warehouses, data centers, low-paid annotation, creative work, and the collection of human behavior. AI is manufactured through material and social systems. Its apparent independence rests on a large infrastructure of people and resources that remains present even when an interface feels effortless.
Her challenge to the word “intelligent” is equally precise. Contemporary models can classify, predict, generate, and optimize with remarkable effectiveness. Yet those achievements do not necessarily amount to general understanding. Systems recognize statistical regularities in the material they receive and operate within objectives selected by designers. They may produce fluent language while inventing facts, or perform well on a benchmark while failing in an unfamiliar setting. Calling every capability “intelligence” can blur important distinctions among pattern recognition, reasoning, agency, consciousness, and judgment. Better descriptions make evaluation clearer and expectations more realistic.
The quote also makes power visible. Building large models requires access to data, computing infrastructure, expert labor, and capital. Organizations that control those resources can influence which problems attract investment, which languages are supported, and which risks receive attention. Communities may contribute cultural material or absorb environmental costs without sharing equally in the benefits. An AI system therefore cannot be assessed only by looking at its output. Responsible analysis follows the full chain: where inputs came from, who performed hidden work, how energy and water were used, who owns the system, and who can contest its effects.
Crawford’s wording is not a denial that AI systems are consequential. It is a demand to take their consequences more seriously by replacing mystique with a complete account. Once AI is understood as an industrial and political system, different questions become unavoidable. Can a useful result be produced with less computation? Were creators and workers treated fairly? Does the application strengthen public capacity or concentrate control? Are limitations communicated honestly? The future of AI will be shaped not only by algorithms but by choices about infrastructure, labor, law, and ownership. Those choices are human, which means they can be debated and changed. Looking at the whole stack also changes what innovation means. A smaller, well-governed system that meets a real need may be more intelligent as policy than a larger model whose costs and dependencies are ignored. Material accountability belongs inside technical progress, not outside it.
Models depend on minerals, chips, electricity, water, data centers, logistics, and human labor across a long supply chain.
The polished interface hides this infrastructure, but environmental and labor costs remain part of the system.
Prediction and generation can be powerful without proving general understanding, intention, or consciousness.
Precise language helps users match trust to demonstrated capability instead of to an evocative marketing label.
The institutions that control data, compute, and deployment can shape whose problems are addressed and whose interests are protected.
Accountability requires examining ownership, labor, resource use, and the ability of affected communities to challenge outcomes.
Explore more "Quotes of The Day"
Continue exploring how writers, thinkers, artists, and advocates have described the choices that shape an everyday life.