Can AI lead us from industrial to ecological intelligence?

Artificial intelligence cannot make civilisation ecological by itself. But it may, for the first time, give humanity the capacity to perceive and coordinate society at the scale and speed of living systems.

Artificial intelligence did not arrive as an environmental movement. It came wrapped in the language of speed, productivity and competition. Its physical foundations – data centres, semiconductor supply chains, cooling systems, and mineral-intensive hardware – belong unmistakably to the industrial world.

Yet AI may also become one of the pathways through which humanity moves beyond that world.

The claim appears self-contradictory. The International Energy Agency (IEA) projects that global data-centre electricity consumption could more than double to 945 terawatt-hours by 2030, with AI the most important driver.

The United Nations Environment Programme warns that AI’s lifecycle encompasses mineral extraction, emissions, water consumption and electronic waste.

How can such a resource-intensive technology become a pathway toward ecological civilisation?

AI is not inherently ecological. It is a civilisational amplifier. Attached to an extractive economy, it will accelerate extraction. Directed towards regeneration, it could help societies understand and coordinate environmental relationships previously too complex or fast-moving for human institutions to manage.

Its distinctive ecological promise is not that AI is ‘green’. It is that AI can shorten the distance between an ecological signal and an institutional response.

The decisive question, therefore, is not whether AI is environmentally good or bad. It is whether we can change the purpose for which intelligence itself is organised.

Why the old pathway was not enough

Past environmental movements made ecological damage politically visible. They gave us protected areas, pollution controls, climate agreements, and the idea of sustainable development.

But most environmental governance remained corrective. Industrial society extracted, manufactured, and consumed; environmental policy then measured the consequences and attempted to regulate, mitigate, or compensate for them. Nature was protected in one department while economic growth was pursued in another.

Its characteristic sequence was extraction damage, measurement, mitigation and repair.

Even the green transition often retains this structure. Petrol vehicles become electric, but dependence on automobiles remains. Electricity becomes renewable, yet ever-expanding demand is rarely questioned.

These interventions are necessary, but substitution is not transformation. A society covered in solar panels can remain extractive. A carbon-neutral economy can remain unequal. A digitally optimised city can still treat forests, rivers, animals and vulnerable communities as externalities.

Ecological civilisation begins when ecological relationships cease to be confined to a single policy field and become society’s operating logic. The economy is not a system alongside nature but a human arrangement nested within the biosphere.

This changes the meaning of progress. Gross domestic product records even the repair of destruction. It cannot tell us whether soil is regenerating, whether a river remains alive, or whether current prosperity diminishes future choices.

The Sustainable Development Goals shifted global thinking toward interdependence. Yet climate, biodiversity, food, energy and digital policies remain institutionally siloed. The ecological crisis is not only a failure of commitment; it is also a failure of coordination.

AI matters because it is potentially a coordination technology. Its importance extends far beyond using a chatbot to compose an environmental report. It offers the possibility of institutions that can receive ecological feedback and change course while action is still possible.

From correction to anticipation

The ecological crisis is partly a crisis of time. Governments act after forests burn, fisheries collapse, or floods displace communities. Knowledge arrives fragmented across disciplines and agencies; by the time institutions assemble it, the system has already changed.

AI introduces a different sequence.

Observation, prediction, prevention, adaptation and learning.

GenCast produced 15-day probabilistic weather forecasts that outperformed the European Centre for Medium-Range Weather Forecasts’ leading operational ensemble system across most variables and lead times tested. Such forecasts can support evacuation, agriculture, renewable-energy management and disaster preparedness.

Flood prediction is even more socially consequential. Many vulnerable communities are located in river basins with few gauges and limited forecasting capacity. A global study found that an AI system could generate useful forecasts for ungauged watersheds and, in many cases, match or exceed the reliability of a major conventional global forecasting.

This represents a philosophical change as much as a computational one. Earlier systems asked how society should respond to disaster. AI-enabled systems increasingly ask whether their human consequences can be anticipated before they become irreversible.

Prediction alone, however, saves no one. Warnings must be trusted, communicated and linked to institutions capable of acting. Historical data may also become less reliable as climate change produces unfamiliar conditions, while poorly monitored regions may receive weaker forecasts despite the appearance of global coverage.

The test is not whether AI produces an answer everywhere. It is whether uncertainty remains visible, local experts can challenge the output, and public institutions can act on it.

Making hidden systems visible

The ecological crisis is also a crisis of perception. Industrial economies push consequences out of sight: atmospheric carbon, extraction at sea, remote biodiversity loss, and exported electronic waste.

AI, combined with satellite imagery and sensors, can bring some of these hidden activities into public view.

A 2024 Nature study led by Global Fishing Watch analysed two petabytes of satellite data to map industrial activity at sea. It found that 72% to 76% of industrial fishing vessels were absent from public tracking systems, with much of the untracked activity concentrated around South and Southeast Asia and Africa.

This does not end illegal or unsustainable fishing. It changes the politics of invisibility. What was once denied because it was hard to observe becomes available for investigation and public scrutiny.

The steam engine expanded power, electricity extended reach, and the internet connected communication. AI can extend pattern recognition across systems too large for a single institution to grasp.

Yet data cannot simply ‘speak for nature’. Sensors reflect what humans choose to measure, and models inherit the priorities and gaps embedded in their data. A machine may process more hours of forest sound than any biologist could, but it does not therefore understand what the forest means.

Ecological intelligence must integrate computational visibility with field science, community memory and Indigenous knowledge. AI may reveal a pattern; societies must still decide what matters, whose knowledge counts and what obligations arise.

From efficiency to ecological coordination

Industrial civilisation standardised production on an enormous scale while assuming it could expand regardless of local limits.

Living systems operate differently. Energy supply varies with sunlight and wind. Water availability varies by place and season. Crops respond to local soils. Transport demand shifts by the hour. Ecological governance requires continuous adjustment rather than static plans.

AI can coordinate renewable generation with demand, detect methane leaks, anticipate equipment failures, optimise building energy use, reduce unnecessary logistics, and trace material flows.

The IEA estimates that widespread adoption of existing AI applications could reduce carbon dioxide emissions by around 1.4 billion tonnes in 2035 – approximately 5% of projected energy-related emissions.

Nothing guarantees this outcome. Infrastructure, data and skills gaps, regulatory barriers, and rebound effects could leave the benefit marginal. Technical potential becomes ecological benefit only through purposeful institutions and incentives.

AI can also support a circular economy through product design, predictive maintenance, material identification, and waste sorting. But circularity cannot mean recycling poorly designed products more efficiently.

If products remain intentionally short-lived and difficult to repair, AI merely sorts the consequences of poor design. Ecological civilisation begins upstream – with durability, repairability, sufficiency, and waste prevention.

This is the critical difference from the past. Traditional environmental policy placed limits on the industrial system, whereas ‘Ecological AI’ could embed environmental feedback into everyday decisions.

But efficiency is not civilisation.

If AI lowers production costs and companies produce twice as much, total resource use may rise. If autonomous vehicles make travel easier but encourage longer trips, energy consumption may increase. If data centres become more efficient while society deploys exponentially more models, the environmental gain disappears.

AI can optimise an unsustainable civilisation without changing its destination. The most efficient route toward the wrong future remains the wrong future.

AI’s material and political contradiction

Any credible case for ecological AI must recognise that the cloud is material: buildings, power stations, cooling systems, cables, water, and mined resources.

US data centres consumed about 176 terawatt-hours of electricity in 2023, representing 4.4% of national use. Lawrence Berkeley National Laboratory estimates that this could rise to between 325 and 580 terawatt-hours by 2028 – equivalent to 6.7% to 12% of United States electricity consumption.

These figures do not prove AI is unjustifiable; they prove that its uses require justification. Society should distinguish flood prediction and methane detection from disposable content, intensified advertising, and speculative consumption.

Energy is only part of the footprint. Chips require minerals and water-intensive manufacturing; data centres can strain local water systems; rapid hardware turnover creates waste. Communities supplying resources or hosting infrastructure may be far removed from those enjoying AI’s benefits.

This can reproduce a colonial pattern: environmental costs are localised while digital value is concentrated elsewhere.

Ownership introduces a second contradiction. As environmental perception becomes computational, knowledge of the planet may depend on privately controlled satellites, models, and cloud platforms.

Countries may contribute ecological data while remaining dependent on systems designed elsewhere. This is not merely unequal access to tools. It is an unequal power to define environmental reality.

Communities should therefore participate in deciding what data are collected, how they are interpreted, and who benefits. Indigenous and local knowledge holders must be partners with authority, not sources from whom information is extracted.

AI does not become ecological merely because its subject is nature. It becomes ecological when the relationships through which it is developed are reciprocal, accountable, and just.

From artificial to ecological intelligence

Artificial intelligence identifies patterns and optimises objectives. Ecological intelligence asks whether the objectives themselves sustain life. It brings computational capacity into dialogue with ethics, planetary boundaries, cultural knowledge and intergenerational responsibility.

An ecological approach to AI requires five commitments: disclosure of energy, water, and material impacts; evaluation based on total ecological outcomes rather than efficiency per transaction; public access to essential climate and biodiversity infrastructure; integration of computational evidence with field and community knowledge; and democratic participation in deciding which problems merit AI resources.

These are not ethical questions to be addressed after the fact. They are design principles. Without them, ‘AI for sustainability’ risks becoming another green label for technological expansion.

The university’s civilisational responsibility

Universities educate AI and sustainability specialists as if preparing for different futures. Computer scientists optimise systems without always asking what deserves optimisation; environmental students may not learn how algorithms organise energy and agriculture; humanities scholars are often invited only after systems have been built.

Ecological civilisation cannot emerge from this intellectual separation.

Every university AI strategy needs an ecological account: computing’s material costs, the problem’s public value, and the distribution of benefits and harms. Campuses can use buildings, energy, food, transportation, and waste as living laboratories.

Procurement matters too. Universities purchase cloud and AI services without routinely requiring comparable information on energy sources, water stress, hardware lifecycles, or emissions.

Sustainability commitments remain in one office while computing contracts are signed in another – the same institutional separation that weakened earlier environmental policy.

Universities should require lifecycle disclosures from technology providers, select the smallest model adequate for a task, and extend research ethics to address material environmental impacts.

Environmental literacy should be integrated into computing education, and AI literacy should be integrated into environmental education.

Most importantly, universities must preserve the intellectual capacity to say no. Innovation is not the insertion of AI into every available space. Sometimes the smart choice is a simpler model, a stronger public institution, a local researcher – or no automation at all.

The defining graduate of the AI era should not merely know how to work with intelligent machines. They should know how to judge which forms of intelligence a living world requires.

A civilisational choice

AI is not an ecological civilisation. Nor can technological innovation replace political courage, reduced consumption, institutional reform, or changed values.

But AI may narrow the gap between ecological change and human understanding. It can move society from delayed repair to anticipation, from hidden extraction to visibility, from standardised production to adaptive coordination, and from fragmented knowledge to systems awareness.

The pathway differs from the past because it does not merely ask society to become less harmful. Conservation places boundaries around nature. Regulation placed limits on pollution. Sustainable development attempted to balance competing priorities. Ecological intelligence can embed environmental feedback into the processes by which priorities are formed.

If AI serves an economy that defines progress as unlimited extraction, it will intensify the crisis with extraordinary efficiency. If governed as public and ecological infrastructure, it may help institutions learn from the living systems they inhabit.

The question is no longer whether machines will become intelligent enough to manage the world. It is whether humanity will become wise enough to decide what intelligence – and civilisation – are for.

James Yoonil Auh is a professor at Kyung Hee Cyber University in South Korea, where he teaches and conducts research on artificial intelligence and global learning systems in higher education. Beyond academia, he has led international education and cultural exchange initiatives across four continents, focusing on sustainability, artificial intelligence, and cross-border collaboration in higher education.

This article is a commentary. Commentary articles are the opinions of the author and do not necessarily reflect the views of University World News.