Overshoot

In the AI Story

The distinction between maximum power and overshoot is Odum's most consequential refinement of Lotka's original principle. Maximum power is sustainable; overshoot is not. From inside the system, they look identical — the predator population booms, the organization ships features, the builder produces at unprecedented rates. From outside, and over longer timescales, they diverge radically. Maximum power sustains itself. Overshoot crashes.

The invisibility of overshoot during the overshoot period is its most dangerous feature. Ecological data shows that predator populations continue to grow for some time after the prey population has begun to decline, because the predators are consuming the accumulated reserve of prey biomass. Similarly, knowledge workers continue to produce at high rates even as their cognitive reserves deplete, because the output draws on stores accumulated before the overshoot began. The warning signs appear late — after the reserves are significantly drawn down, sometimes past the point of recovery.

At the individual level, the Berkeley researchers measured overshoot signatures in AI-augmented workers: increased exhaustion, diminished satisfaction, eroded empathy, the flat affect of nervous systems running beyond sustainable rates. These are not character failures; they are the predictable signatures of systems consuming cognitive energy faster than replenishment.

At the civilizational level, the concern is whether intellectual, institutional, and ecological stores are being drawn down faster than they can be regenerated. The prosperous way down is Odum's proposal for avoiding catastrophic overshoot through deliberate deceleration — an approach rejected, so far, by the incentives governing the AI growth phase.

Origin

The concept has roots in population ecology — Georgii Frantsevich Gause's 1934 experiments, Paul Ehrlich's 1968 work, and William Catton's 1980 Overshoot: The Ecological Basis of Revolutionary Change. Odum integrated the concept into systems ecology as the failure mode of maximum-power configurations that lose their feedback loops.

The framework's extension from ecological systems to economic and cognitive systems follows Odum's universalist method: the same laws govern all self-organizing systems that sustain themselves through continuous energy processing.

Key Ideas

Invisible during overshoot. The system's metrics look healthy precisely when the depletion is most severe.

Feedback loops fail. The mechanisms that should moderate consumption are overwhelmed or bypassed during overshoot.

Reserves mask depletion. Accumulated stores sustain output after the sustainable rate has been exceeded.

Crash follows abundance. Collapse occurs after the reserves are exhausted, often suddenly from the perspective of participants.

Prevention requires measurement. Distinguishing maximum power from overshoot requires measuring the energy base, not just the output.

Debates & Critiques

Whether AI deployment is currently in overshoot is contested. Proponents argue that efficiency gains and renewable energy deployment will close the gap. Critics, including the Odum framework applied here, argue that the maximum power principle predicts demand will grow at least as fast as efficiency gains, and that the measurement of intellectual and institutional storage — not just electrical efficiency — is the relevant test.