Use versus Exploitation (Berry)

In the AI Story

Berry developed the use/exploitation distinction through direct observation of two farming practices on adjacent Kentucky hillsides. One farmer—Berry's neighbor—tended the land with rotation, rest, compost, minimal tillage. The soil grew darker, richer, more productive over decades. The other farmer maximized yield through monoculture, heavy fertilizer, annual plowing that left soil exposed to erosion. Yields were higher initially. Within fifteen years, topsoil had eroded to bedrock in places, organic matter had declined measurably, the land required increasing chemical inputs to produce declining outputs. Same climate, same slope, same initial soil conditions. The difference was the relationship between the farmer and the land. One relationship was use; the other was exploitation. The exploited land did not recover within the exploiting farmer's lifetime.

The distinction maps onto AI-augmented work with precision that should disturb anyone paying attention. When Edo Segal's engineer uses Claude to handle "plumbing"—dependency management, configuration—and loses ten minutes of formative struggle embedded in four hours of tedious work, the loss is not of the tedium (which was genuinely unproductive) but of the unexpected moments when the system revealed connections no documentation conveyed. Those moments were use—direct encounter with the material that deposited understanding. Their elimination is exploitation—taking the output (working code) without investing in the capacity (embodied knowledge of how systems behave) that makes future outputs sustainable. The engineer's confidence in architectural decisions declined months later. The soil had thinned.

Berry insists the distinction is not merely moral (though it is moral) but practical and empirical. Exploitation produces higher short-term yields and lower long-term capacity. The timeline of degradation is what makes exploitation economically viable—you can profit for years, sometimes decades, before consequences arrive. But consequences always arrive. Exploited soil compacts, erodes, loses fertility. Exploited codebases accumulate technical debt, become unmaintainable, collapse under the weight of their own complexity. Exploited workers burn out, leave, take their irreplaceable tacit knowledge with them. Exploited attention—the continuous partial attention that AI-saturated environments produce—degrades the capacity for deep reading, sustained thought, genuine presence to other people. The degradation is invisible to productivity metrics. It is visible to anyone willing to observe what happens to the practitioner over time.

The most uncomfortable application is to AI collaboration itself. Segal's partnership with Claude in writing The Orange Pill occupies an ambiguous position: Is the collaboration use (two different forms of intelligence producing insights neither could achieve alone) or exploitation (Segal extracting articulation capacity from Claude without understanding the mechanisms producing it)? Berry's framework suggests the answer depends on whether the collaboration deepened Segal's understanding of his own thinking or merely provided a more efficient way to produce output. Segal's confession that he "could not tell whether I actually believed the argument or whether I just liked how it sounded" after Claude generated a passage suggests moments of exploitation—taking language without earning it. His practice of deleting that passage and writing by hand until he found his own version suggests use—employing the tool to clarify what he genuinely thought rather than to bypass the thinking.

Origin

Berry first articulated the use/exploitation distinction in The Unsettling of America (1977), where he diagnosed the industrial agricultural system as organized around exploitation: "The industrial economy is based on the assumption that it is possible to take without giving back, that the creation is a sort of mining operation." The sentence was not metaphorical. It described a material reality Berry had observed: industrial farms treating soil as a non-renewable resource, extracting nutrients faster than they could be replenished, producing record yields for decades before the soil gave out. The Green Revolution's temporary success—which fed billions and won Norman Borlaug a Nobel Prize—was purchased at the cost of soil degradation, aquifer depletion, and chemical dependency that subsequent generations are inheriting as a permanent crisis. Berry's framework predicted this outcome in 1977, not through prophetic insight but through observing what happens to any system when extraction exceeds investment in the system's productive capacity.

The distinction gained new urgency in Berry's 2012 Jefferson Lecture, where he warned that human intelligence "could get us into trouble that it could not get us out of" if wielded without humility about its limits. The warning applied to genetic engineering, industrial agriculture, and financial systems—and applies with equal force to AI, which operates on the same extractive logic: claiming humanity's accumulated knowledge as free raw material, processing it into proprietary capability, distributing the outputs while concentrating the gains. Berry would note that this is textbook exploitation—taking without giving back, mining the intelligence commons without investing in the conditions that produced it.

Key Ideas

Exploitation produces higher short-term yields. The economic logic favoring exploitation over use is real—mined soil produces more in year one, AI-assisted developers ship more in quarter one—making the choice to tend rather than mine a choice against immediate competitive advantage.

Use requires knowledge exploitation does not. Tending a domain requires understanding its nature, limits, needs; exploitation requires only the capability to extract—the farmer who strip-mines soil needs no knowledge of soil science, the developer who uses AI without understanding needs no knowledge of the implementation.

The timeline of degradation conceals the cost. Consequences of exploitation arrive on timescales (years, decades) longer than the planning horizons (quarters, product cycles) governing most organizational decisions, making the degradation invisible to the decision-makers until it becomes irreversible.

Measurement systems hide exploitation. Productivity metrics, quarterly earnings, benchmark scores measure extraction; they systematically fail to measure the domain's long-term capacity to sustain extraction, making exploitation appear successful until the capacity collapses.

Use is a practice, not a sentiment. Berry's "care" is not warm feeling but competence—the specific knowledge of what a domain needs, the willingness to provide it, the discipline to resist extraction when extraction would be profitable, built through sustained attentive presence to the thing being tended.