The Anthropocene Experiment

The fossil record is often invoked as consolation. Earth has survived worse. It has endured asteroid strikes, continents drowned beneath lava, oceans robbed of oxygen, and climates altered so violently that entire evolutionary worlds disappeared. But the reassurance conceals a fatal confusion. Earth’s survival is not the same thing as the survival of its creatures, its forests, its reefs, its civilizations, or the particular conditions that make a planet inhabitable to us.

The lesson of mass extinction is not that life is invulnerable. It is that living systems can collapse with terrible speed once several pressures converge.

Elizabeth Kolbert’s The Sixth Extinction and Peter Brannen’s The Ends of the World approach this fact from opposite ends of time. Kolbert enters the present tense of extinction: amphibians undone by disease, island species overwhelmed by arrivals they never evolved to resist, forests diminished into fragments, and oceans quietly remade by the carbon humans have placed in the air. Brannen moves through older catastrophes: the end-Ordovician freeze, the end-Permian inferno, and the impact winter that closed the age of non-avian dinosaurs. Together, their books make a single grim argument. Extinction is not always a visible catastrophe arriving from the sky. It can be the accumulation of ordinary processes—heat, altered chemistry, disrupted habitat, disease, starvation, displacement—until the world becomes unrecognizable to the organisms that evolved within it.

We are now conducting that experiment deliberately, if not always consciously. We burn fossil carbon that took millions of years to bury. We clear and divide habitats faster than species can relocate. We drain aquifers, reroute rivers, trawl oceans, simplify fields into monocultures, transport pathogens and invasive species through global trade, and disperse chemical residues through soils, water, and food webs. We call these separate problems: climate change, biodiversity loss, pollution, water scarcity, food insecurity, resource depletion. But the separations exist mostly in the filing cabinets of governments, corporations, and universities. In the living world, they are one event.

That event is the polycrisis.

No Single Killer

A polycrisis is not simply a crowd of bad events occurring at once. It is a condition in which crises share causes, amplify one another, and reduce the ability of societies and ecosystems to respond. Climate disruption worsens drought, flood, wildfire, disease, crop failure, displacement, infrastructure breakdown, and political instability. Habitat fragmentation prevents species from making the migrations by which they once escaped changing climates. Pollution weakens organisms already under thermal stress. Inequality determines who can purchase water, insurance, air conditioning, medicine, relocation, and private security—and who remains exposed to the damage.

The sixth extinction is the biological face of the polycrisis. It is where the abstractions of industrial civilization become visible as absences: the bird no longer heard, the amphibian no longer found, the coral reef no longer built, the insect population reduced beyond ordinary notice, the forest still green but emptied of its intricate inhabitants.

The word “mass” can mislead if it is used carelessly. The present crisis has not yet reached the completed scale of the Big Five, generally defined as the loss of roughly three-quarters or more of species over a geologically brief interval. But extinction is not measured only by the final body count. It is measured by rate, direction, and the narrowing of the future. IPBES reports that the global rate of species extinction is already tens to hundreds of times higher than the average over the past 10 million years, and that it is accelerating. Around one million animal and plant species are now threatened with extinction, many within decades. The IUCN Red List, which assesses only a portion of the world’s species, lists more than 47,000 threatened species—about 28% of those assessed.

The comparison with the ancient extinctions is therefore not a claim that the work is already finished. It is more disturbing than that. The mechanisms are active, the losses are accelerating, and the number of species pushed toward disappearance is vast. Even a deliberately conservative analysis of modern vertebrate loss concluded that extinctions over the last century occurred at rates up to one hundred times above background, and that losses recorded in a century would otherwise have taken hundreds to thousands of years. We are not looking back on a completed extinction horizon. We are watching one gather speed.

The extinction crisis is not a collection of independent environmental problems. The IPBES Global Assessment identifies five major direct drivers of biodiversity loss: land- and sea-use change, direct exploitation of organisms, climate change, pollution, and invasive alien species. Its more recent assessments show that biodiversity loss is entangled with water, food, health, governance, inequality, production, consumption, and technological development. The IPBES Nexus Assessment and IPBES Transformative Change Assessment make clear that these crises share drivers, feedbacks, and trade-offs rather than existing in separate policy silos. These are not five unrelated misfortunes. They are the ecological outputs of an economy organized to convert land, water, minerals, animals, forests, and atmospheric stability into temporary economic value. The danger lies not only in losses already recorded, but in the vast number of species whose habitats, food webs, climatic ranges, and reproductive conditions are being degraded faster than recovery can occur.

The phrase “humanity is causing extinction” is true only from a distance so great that it becomes evasive. Humanity is not a single actor. A family facing a failed harvest, a community living beside an industrial sacrifice zone, an Indigenous nation defending its water from a mine, and the executives of fossil-fuel, shipping, mining, agribusiness, and technology companies do not bear equal responsibility. Nor do they possess equal power to determine what happens next.

The principal agent of the present extinction crisis is not humanity in the abstract, but a global industrial order organized around concentrated wealth, rising material throughput, fossil energy, extractive frontiers, unequal consumption, and institutions that routinely shift ecological costs onto the public, politically marginalized communities, other species, and future generations.

This is why extinction cannot be solved merely by asking individuals to become more responsible consumers while the systems governing energy, food, transportation, investment, housing, and land use remain committed to expansion. A society can recycle, compost, buy efficient appliances, and still clear forests for commodity crops, subsidize fossil extraction, expand highways, strip-mine deserts, and sacrifice watersheds to the imperative of growth. The problem is not that civilization has received insufficient information. It is that the institutions holding power continue to treat ecological limits as obstacles to be overcome rather than conditions of survival.

The Archive of Dead Worlds

Brannen’s account of past mass extinctions matters because it refuses the comfort of a single explanation. The end-Permian extinction—the largest of the Big Five—is widely associated with massive volcanism and rapid carbon release, followed by extreme warming, ocean acidification, deoxygenation, and possibly toxic atmospheric effects. So many overlapping mechanisms may have contributed that one paleontologist has described the disaster as a “Murder on the Orient Express” event: everybody may have done it.

The image is almost too exact for the present.

There is no single assassin in the sixth extinction. No one villain stands alone in a laboratory, a boardroom, a mine, or an oil field. The killers arrive together: fossil combustion, industrial agriculture, trawling, logging, roads, dams, plastics, pesticides, heat waves, droughts, invasive species, wildlife disease, chemical runoff, financial speculation, planned obsolescence, and the permanent political demand that every year’s economy be larger than the last.

Each force arrives with paperwork. Each has an economic rationale. Each is presented as necessary, efficient, competitive, temporary, regrettable but unavoidable.

The fossil record suggests that the most devastating changes in Earth’s history were often compounded. A disrupted atmosphere altered the climate. Altered climate changed ocean circulation. Changed circulation reduced oxygen. Reduced oxygen reorganized microbial life. Acidification undermined shell-builders and reef-makers. Habitat changed faster than organisms could move or adapt. Food webs collapsed from the bottom upward. Life did not fail because one stressor was unpleasant. It failed because several stressors destroyed the margins within which adaptation remained possible.

The modern world differs in one decisive way. The end-Permian catastrophe did not have highways, industrial fishing fleets, chemical-intensive monocultures, derivatives markets, global tourism, data centers, advertising systems, military supply chains, or machine-learning models. It did not have an organism capable of recognizing the mechanisms of catastrophe and then building institutions dedicated to accelerating them.

That organism is industrial civilization.

The Ocean Remembers First

The ocean is one of the clearest places to see the connection between deep time and the present emergency. In The Sixth Extinction, Kolbert visits volcanic carbon-dioxide vents near Castello Aragonese, Italy, where seawater grows increasingly acidic close to bubbles rising from the seafloor. The result is not a mere adjustment in a scientific graph. It is a thinning of life: shell-building organisms decline or disappear; mussels, barnacles, corals, snails, and calcifying algae struggle where altered chemistry makes their basic biological architecture more difficult to maintain.

The future ocean is not hidden in a distant projection. In places like this, it is visible in miniature.

Carbon dioxide does not disappear when it leaves a smokestack, tailpipe, or gas turbine. Some remains in the atmosphere and warms the planet. Some enters the ocean and changes seawater chemistry. For organisms that depend on carbonate to build shells, skeletons, and reefs, this is not a secondary inconvenience. It is an assault on the material basis of their existence. The ocean absorbs the costs of industrial civilization until that absorption becomes another form of damage.

Brannen’s account of the end-Permian disaster offers the unnerving deep-time echo. Vast carbon injections helped drive extreme warming and destabilize ocean chemistry. Acidification and oxygen loss appear among the major proposed mechanisms by which marine ecosystems lost their complexity. The relevance is not that the twenty-first century will reproduce the Permian exactly. It will not. Continents, ecosystems, atmospheric conditions, and timescales differ. The relevance is that rapid carbon release has before transformed the ocean from a cradle of diversity into a medium hostile to much of the life it once supported.

The key word is rapid.

Earth has slow stabilizing processes. Over immense spans, rock weathering can draw carbon from the atmosphere and carry it toward the sea. Sediment can bury carbon. Evolution can eventually produce organisms suited to altered conditions. But these processes operate on timescales indifferent to the lifespan of a coral reef, a fishery, a coastal city, or a human generation. A civilization that burns geological carbon in centuries is not merely adding carbon to a system. It is compressing deep time into an emergency.

The ocean remembers first because it has nowhere else to put what we give it.

Extinction by Normal Means

The most frightening feature of the sixth extinction may be its ordinariness. It does not require an obvious apocalypse. It can be produced by normal operations.

A forest road divides habitat. A species that once moved across a continuous landscape now encounters vehicles, fences, farms, subdivisions, and rising heat. A wetland becomes a logistics park. A river is dammed, diverted, warmed, or depleted. A cargo vessel transports an unfamiliar organism across an ocean. A pathogen enters a population with no defense against it. A reef experiences another bleaching event, then another, until recovery becomes less likely than replacement by algae. A fish population is managed for annual yield until the conditions that allow it to replenish disappear.

No isolated decision seems to contain the end of a world. That is how a world ends while appearing administratively normal.

Kolbert’s great strength is her attention to loss at the scale of organisms and places. Extinction is not a distant geological category. It occurs through changed relationships: species encountering new competitors, new diseases, new temperatures, altered ocean chemistry, and landscapes broken into pieces too small or isolated for survival.

The problem is compounded by shifting baselines. Each generation inherits a diminished world and calls it normal. A child born after the disappearance of a wetland does not experience the wetland’s loss as loss. A forest emptied of birds and large animals can still look green from the highway. A river reduced to a managed channel can still be called a river. A summer nearly absent of insects can appear merely quiet.

Extinction succeeds twice. First it removes creatures from the world. Then it removes the memory that their presence was ever ordinary.

This is why conservation must mean more than preserving a remainder. It must mean defending the conditions of complexity itself: intact habitat, migration routes, reproductive cycles, clean water, living soils, ocean chemistry, climatic stability, and enough room for other species to remain more than decorative remnants in a managed human landscape.

The newest systems do not interrupt this ordinary machinery of loss. They promise to make it more efficient.

The Machine in the Extinction Event

Artificial intelligence does not need to become conscious, malicious, or superhuman to become dangerous. Direct loss-of-control risks remain uncertain and contested; the more immediate danger is that AI will be deployed at scale within institutions already rewarded for accelerating ecological and social breakdown.

The popular language of AI encourages a fantasy of disembodiment: intelligence in the cloud, answers at the speed of thought, a digital layer floating above the material world. But the cloud is not weightless. It is an industrial landscape of data centers, power plants, transmission corridors, cooling systems, chip factories, mines, freight networks, water withdrawals, and enormous claims on electricity. Its apparent immateriality is an effect of distance. The consequences are moved out of sight, often onto communities and landscapes with the least power to refuse them.

The International Energy Agency projects that global data-center electricity consumption will rise from roughly 415 terawatt-hours in 2024 to about 945 terawatt-hours by 2030, with AI a major driver of that expansion. It also expects natural gas and coal together to meet more than 40% of the additional electricity demand from data centers through 2030. IEA: Energy Demand from AI and IEA: Energy Supply for AI. The ecological question is therefore not whether machines can think. It is what kind of physical world must be consumed to make their thinking possible.

AI may help detect methane leaks, map wildfire risk, monitor deforestation, forecast extreme weather, identify endangered species, and manage electricity systems. Such uses matter only when they are joined to institutions willing to act on what they reveal. The same capacity for prediction and optimization can also serve other ends: finding new mineral deposits, coordinating extraction, targeting consumption, pricing climate risk, automating layoffs, managing supply chains, expanding surveillance, and accelerating military decision-making.

What changes is not the moral character of the system, but its speed, reach, and ability to hide political decisions behind technical procedures.

That distinction becomes dangerous under the competitive logic of the prisoner’s dilemma. Every firm may understand that unlimited computation, energy consumption, extraction, and automation deepen collective risk. Every state may understand that military competition and data-center expansion can worsen climate pressure, resource conflict, and instability. Yet each fears that restraint will merely grant an advantage to a less restrained rival. The result is a race in which everyone becomes more powerful in the short term and less secure in the long term.

Build the larger model. Secure the mineral supply. Expand the data center. Automate before being automated. Deploy before competitors do.

What is rational for each competitor becomes ruinous for the world.

This is how AI enters the sixth extinction. Not necessarily as an autonomous executioner, but as an accelerator of the systems already making Earth less habitable: fossil energy, material throughput, habitat conversion, extraction, militarization, and unequal adaptation. It can make a society more precise in its understanding of the damage without making it more willing to stop. It can calculate that an aquifer is failing, a coast is flooding, a forest is burning, or a species is disappearing—and then help decide whose water is cut off, whose home becomes uninsurable, which community is abandoned, and which profitable activity continues.

Artificial intelligence can increase the resolution of perception without increasing the capacity for restraint.

The danger is not that the machine will necessarily destroy humanity on its own. The danger is that humanity may use the machine to become more efficient at destroying the systems that make human life possible.

The Management of Scarcity

The most politically dangerous possibility is not that AI will create ecological breakdown. Ecological breakdown is already underway. The danger is that AI may turn its consequences into an increasingly efficient system of allocation, exclusion, and abandonment.

The first form of collapse is often administrative. A service is reduced, a claim is denied, a repair is postponed, a rate rises, a public system is told to do more with less. No siren announces the transfer of risk from institutions to households. It arrives as paperwork. People are asked to adapt one invoice, one water restriction, one cancelled policy, one delayed repair at a time, until security is redefined as the private ability to pay for failures that public systems have chosen not to prevent.

As climate disruption worsens, the question of who receives protection becomes unavoidable. Who receives cooling during lethal heat? Who can insure a home? Who is relocated after fire, flood, storm, or sea-level rise? Who receives scarce water? Which crops are subsidized? Which communities are abandoned? Who is classified as a security threat when drought, hunger, and war produce migration?

These are political questions. AI can make them faster, more opaque, and more scalable.

A government could use automated systems to locate vulnerable households and direct aid effectively. It could also use them to deny benefits, monitor populations, predict migration, target dissent, automate border enforcement, or convert human survival into a risk score. Corporations can use AI to reduce certain forms of waste and improve grid management. They can also use it to stimulate consumption, expand fossil extraction, automate layoffs, intensify speculation, and identify new territories of profitable exploitation.

Technology does not resolve the values of the institutions that deploy it. It magnifies them.

This is especially dangerous in a world where ecological strain is readily converted into coercion. That outcome is not mechanically inevitable, but history gives little reason to expect its opposite. Ecological stress does not automatically produce war, authoritarianism, or cruelty; it exposes and intensifies the hierarchies already present. In societies organized around unequal wealth, private property, national rivalry, and the protection of capital, scarcity is rarely shared fairly. It is priced, securitized, and imposed downward. Water becomes an asset, housing becomes an investment vehicle, migration becomes a threat category, and survival becomes something purchased by those able to purchase it.

Under those conditions, artificial intelligence is unlikely to arrive as a neutral instrument of public care. More plausibly, it will be folded into the existing machinery of triage: calculating risk, withdrawing insurance, screening migrants, allocating scarce resources, monitoring dissent, and making abandonment appear as the objective result of a model rather than a political decision. The danger is not that technology will somehow choose injustice on its own. It is that it will make injustice faster, cheaper, more comprehensive, and easier to describe as necessity.

The machine will not have to choose who is sacrificed. The institutions that own it already have.

The Layer We Are Making

Kolbert writes of a future stratigrapher reading the present in the rocks: altered carbon, rearranged species, extinctions, plastics, radionuclides, diverted rivers, transformed sediment, and the strange biological signature of organisms moved, domesticated, favored, and discarded by human activity. The Anthropocene, whatever name future geologists finally give it, will be visible as a rupture.

The future fossil layer will not record our intentions. It will record consequences.

It may show an ocean altered faster than its shell-builders, reef-makers, and food webs could adapt. It may preserve the signature of carbon released from ancient deposits and returned, in an instant of geological time, to the atmosphere. It may mark the disappearance of species that survived ice ages, continental drift, and previous climatic upheavals, only to be undone by roads, monocultures, invasive predators, chemical contamination, warming seas, and an economy that found their continued existence inconvenient.

It may also record the machinery built to administer that loss: metals, plastics, altered landscapes, electronic waste, power infrastructure, and the residues of a civilization that mistook computational power for permission.

Artificial intelligence may not be the author of the sixth extinction. That extinction is already being written by a much older system: industrial growth without ecological restraint. In the absence of democratic and planetary limits—limits the present order has shown little willingness to impose—AI may become the fastest clerk in that system: the instrument that records, prices, predicts, and accelerates not only the disappearance of the living world, but the unraveling of the human world that depends upon it.

The decisive question is not whether our machines become intelligent.

It is whether, at the moment our capacity to calculate became almost limitless, we discovered too late that efficiency is not wisdom—and that a civilization can become brilliant at measuring the world even as it destroys the conditions that make a world worth measuring.