Temporal Undulation

In the AI Story

The AI hyperobject undulates across three temporal scales simultaneously. First, the immediate: nanosecond-to-second computational inference, too fast for perception. The user experiences the response as instantaneous; the computation is invisible. This invisibility contributes to smoothness — the response appears without process, without labor, as if given rather than produced. Second, the biographical: months-to-years restructuring of individual minds through habitual interaction. This is the timescale where the Berkeley researchers documented task seepage, work intensification, attention erosion. Changes are perceptible in retrospect but not in real time. The restructuring accumulates like sediment — continuously, invisibly, with consequences apparent only when the delta has shifted the current.

Third, the generational: decades-to-centuries transformation of civilizational cognitive infrastructure. This timescale exceeds any individual life. Generational effects of AI on cognition are entirely unknown as of this writing — not because researchers lack interest but because the experiment has been running less than a decade. The interaction among these three scales produces dangerous imperceptibility. The immediate scale is too fast. The biographical scale is too slow. The generational scale is too vast. Human perception, calibrated for middle timescales, falls into the gap between all three.

The frog-in-gradually-heating-water parable is apocryphal about frogs but accurate about humans inside temporally undulant hyperobjects. The nervous system detects sudden change (the predator, the crack, the light shift) with extraordinary precision. It cannot detect change accumulating on the biographical timescale. By the time the aggregate is large enough to perceive, the perceiver has been transformed by the accumulation — has become a different entity with different baselines, different expectations, different standards for what counts as normal. The damage may be real, cumulative, profound — and nevertheless invisible at every point along the curve.

This is the scenario haunting honest AI assessment: the possibility that something is being lost slowly and continuously, and that the loss is reshaping the apparatus that would need to detect it, so that by the time the loss is detectable it will no longer register as loss. It will register as the way things have always been. Morton does not flinch from this possibility. The temporal undulation of hyperobjects is a genuine problem — not solvable but inhabitable. The difference between those categories is the difference between the engineering mindset (identify problem, design solution, deploy fix) and the ecological mindset (identify condition, develop practices for living within it, accept that the condition will not resolve within a human lifespan).

Origin

Morton developed temporal undulation from the observation that climate change operates on timescales that make human response structurally difficult. A coal plant emits CO₂ today; the atmospheric concentration peaks in decades; the temperature effects manifest across centuries; the ecological cascade persists for millennia. No human institution has been designed to govern consequences unfolding across that temporal range. The mismatch is not a design flaw. It is an ontological condition.

Applied to AI, temporal undulation explains the gap between rapid deployment (months) and institutional response (years to decades). It explains why individual cognitive restructuring by smooth interfaces is invisible in real time. It explains why generational effects remain unknowable while the perturbation propagates. The scotoma is structural. Human experiential time and hyperobject time are incommensurable, and the incommensurability is the condition, not the problem.

Key Ideas

Three timescales operate simultaneously. Immediate (too fast to perceive), biographical (too slow to detect in real time), generational (too vast to observe within a lifespan).

The mismatch produces a scotoma. Not a failure of attention but an ontological gap — deep time cannot be represented by apparatuses calibrated for experiential time.

Damage accumulates invisibly. Cognitive restructuring by smoothness operates on timescales falling outside perceptual resolution, becoming visible only after the perceiver has been transformed.

No solution state exists. The condition is inhabitable, not solvable; biographical interventions cannot prevent generational transformations.

Awareness without mastery. Thinking temporal undulation produces not control but a different quality of action — provisional, attentive, willing to revise as effects manifest.