Cognitive dependence and the ecological cost of AI

Reducing generative AI to a mere tool obscures two essential questions: its hardware composition and its impact on society. This simplistic view prevents us from seeing the systemic effects that our individual choices cannot control.

Invisible infrastructure and its true cost

Behind every request lies a global network: energy-intensive data centers, rare earth mines, undersea cables, and precarious workers moderating content for a few cents. Referring to it as a "tool" erases this colossal material reality.

Digital extractivism and its externalities

This infrastructure is part of a logic of massive extraction: minerals, energy, human attention. The exponential growth in usage, necessary to make investments profitable, negates any effort at individual restraint.

When adoption creates dependency

Just as the automobile restructured urban space, conversational AI is redefining productivity standards. In education, work, and services, refusing the tool is becoming increasingly impossible without self-exclusion.

Accelerated standardization

Education already trains people in the "art of speed" to respond to a supposedly saturated AI market. The need thus arises from its own prophecy, transforming individual choice into a mandatory collective norm.

Strategic double-speak

Industry players are caught between two contradictory narratives: AI is seen as a commonplace tool requiring little regulation, but also as an existential risk justifying centralized governance. This contradiction serves the same purpose: to accelerate adoption while simultaneously controlling power.

Programmed cognitive atrophy

Delegating writing to algorithms bypasses the process of clarifying thought. Writing is not an obstacle, but the laboratory where ideas are built. Automating this step atrophies irreplaceable skills of synthesis and argumentation.

The risk of emotional dependency

Virtual companions, designed to maintain unlimited and empathetic conversation, create reward loops similar to social networks. They simultaneously collect unprecedented intimacy data, making any turning back more difficult.

Towards systemic responsibility

Individual vigilance is not enough. A proportionate framework must include: energy sobriety, traceability of supply chains, rights of click workers, data protection, the right to a non-algorithmic alternative, and anti-monopoly regulation.

Reinventing digital culture

Beyond regulation, we need to transform daily practices: preserve unassisted writing in schools, value human synthesis in business, integrate frugality into software design with session limits and energy transparency.

A global ethic of care

This approach is not technophobia, but an ethic that cares for resources, workers, minds, and institutions. Slowing down does not mean rejecting innovation, but rather reintegrating it into a democratic framework.

The essential question

What kind of world do we want to build by accepting or challenging the social architecture of generative AI? Believing that the answer lies at the individual level overlooks the fact that tools shape structures. Shifting the debate to infrastructure and dependencies allows us to envision a powerful AI, but one anchored to goals other than profit maximization.

To go further: this systemic thinking joins our article on Claude's safety and compliance in business et Our analysis of AI ethics in the face of reality.