Can AI help redesign the technosphere—and so save the planet?

If AI is going to help save the planet, we’re going to have to point it away from the Earth and toward our techno-human selves.

Image: Midjourney

Following on from my post this past week on the new Stockholm Resilience Centre report on AI for a Planet Under Pressure, my good colleague and energy transitions expert Clark Miller provides an alternative and compelling perspective on the intersection between artificial intelligence and our planetary future. Andrew Maynard

As it does each year, the opening of the UN Climate Change Conference (COP 30) in Belém, Brazil, has brought the release of a host of new reports hoping to inject new ideas into the quest to save the planet. So, it’s hardly a surprise that, in a year when AI has dominated headlines, its proponents are now telling us that artificial intelligence will be the magic bullet to do the trick.

Is there anything AI can’t do?

Sorry for the cynicism, but, as I see it, the latest incarnations of the AI planetary savior complex are on the wrong track, including from major players who should very much know better. Both Google, in its new AI for Nature report, and the Stockholm Resilience Centre (SRC), in its AI for a Planet Under Pressure, are stuck in a 50-year-old paradigm for rescuing the planet that has done little to arrest the rising concentrations of greenhouse gas emissions or address the fundamental drivers of planetary change.

If AI is to have a hope of helping humanity to save the planet, we’re going to have to give it a bold new direction.

The Problem

The basic problem is that both Google’s and SRC’s new reports focus almost exclusively on nature. To save the planet, they suggest, we need to better understand Earth systems—fundamental planetary-scale environmental systems like the climate, the biosphere, and biogeochemical cycles—and then better incorporate that information into human decision-making.

AI, they promise, can help us do that.

It’s an old argument. It’s been the premise, for example, since 1990, of the US Global Change Research Program, which has spent $2B per year on Earth Systems science—including the world’s largest fleet of satellite monitoring systems and some of the world’s most sophisticated computational modeling—to inform the US government and industry about planetary dynamics.

Look how far that has gotten us.

To be sure, we’ve learned an enormous amount about planetary systems since Charles Keeling first documented the upward march of carbon dioxide concentrations in the atmosphere in 1958, and the First World Climate Conference in 1979. And those insights have powerfully informed global sustainability policy. But they have also constituted a major distraction.

For a half century of global environmentalism, since well before the 1992 Earth Summit, what has been largely missing from planetary science has been a sufficient and sustained effort to study the primary forces that are actually driving global ecosystems beyond safe planetary boundaries: namely, the planet-wide techno-human systems built to power and provide materials to the world’s economies and societies.

Don’t believe me? Consider this. Since 1990, the world’s most prestigious universities have added hundreds of new degrees, programs, departments, and schools of Earth Systems science. During that time, however, no comparable program has been added to study, observe, or model the structure and dynamics of planetary-scale technological systems or their aggregation into what might be called the “technosphere.”

Not one.

To be sure, almost every major university on the planet has an engineering college. But no engineering college studies global-scale technological systems. They study the core principles of physics, materials, machines, and (sometimes) systems, the fundamentals of the engineering profession, and the basic practices of technological innovation and entrepreneurship.

But they do not study the vast aggregate technological complexes that humans have created at the center of the global economy. They do not analyze how such complexes are designed and operate, or how they interface with human systems. Their classes and research programs don’t cover how humans have used technologies to build an artificial environment for themselves that sits at the interface of the Earth, the biosphere, and the atmosphere, or how those technologies are woven into the planet’s social, economic, cultural, or ecological fabric. They study buildings and construction processes but not the global construction enterprise, its far-flung material supply chains and waste streams, its planet-wide product, or its vital entwinement with the life, work, and wellbeing of humanity or the Earth’s ecological integrity.

At least in the ways that we conventionally define engineering, these aren’t engineering problems. Sure, they are engineering-adjacent. But they are not engineering.

Yet they may be problems that AI could help tackle.

A Glimmer of AI Hope

Here, AI gives me hope from two directions.

The first is observational. AI is now making visible the sheer scale of humanity’s planetary technological enterprise.

I still remember the awe of standing on a hillside overlooking the site of the Manhattan Project’s K-25 uranium separation plant in Oak Ridge, TN. Only a tiny fraction of the former plant remains, but photos showed visitors the vast facility that sprawled across the valley. The scale was mind-boggling: over 1500 acres of interconnected structures that began as the world’s largest building and, by the height of the nuclear arms race in the late 1950s, comprised the Earth’s largest industrial complex.

But how do we visualize—or perhaps more importantly, imagine, and call into our mind so that we can make sense of—the vast technological systems and infrastructures that power the global economy?

To this point, we’ve really only had one good image of the technosphere: NASA’s famous photographs of the Earth lit up at night. The photos don’t show the aggregate, planet-wide techno-human complex, but they do reveal the light it gives off (light pollution that, we now know, comes with lots of consequences, locally and globally, along its lifecycles).

Yet, in the past few weeks, I’ve listened to a number of podcasts with people who’ve visited the construction site for the new Stargate data center in Abilene, TX. To a person, they express the same feeling of astonishment as I did at Oak Ridge for the facility’s almost unimaginable scale. Imagine Wembley Stadium or the Seattle Kingdome. Now imagine 1000 football stadiums densely packed into a 30x30 cluster!

Abilene is one of 5 Stargates being built across the United States. And, if Sam Altman is to be believed, more are coming on every continent.

These are the footprints of humanity’s new ambitions for compute, and are all too visible from space. And they come with a $10T price tag.

It’s no wonder Caterpillar stock keeps going up.

Maybe now we’ll begin to take seriously the project of understanding the aggregate complex of human technological enterprise and its planetary consequences.

And that’s where AI offers, at least potentially, my second hope.

AI as a Technology for Good

If we had to invent a new human discipline of planetary technological systems science, it would take far too long to make much difference in what is now a sprint to meaningfully transform the world’s technological systems toward more sustainable foundations.

And if we humans, on our own, had to come to grips with integrating and synthesizing the massive data required to make sense of how planetary-scale techno-human systems work—how they’re structured and designed, where effective intervention points might be that will tip them from less to more sustainable operating states, what the wider impacts of systems change might be throughout world economies and societies, and how to channel the inevitable human backlash to those impacts in ways that enhance just outcomes and positive transformations rather than catalyzing resistance—I’m not sure that we’d be up to the task.

Here though, maybe AI can help.

I’ve worked for two decades on energy systems transitions, and the task is enormous. To offer one example, we’ve built tremendous capabilities across the sector to model energy technologies and energy systems. But almost none of those models include humans in any meaningful way. Hence, as we reshape energy technologies and systems, the models give no warning of how human-energy relationships will be reshaped. And, thus, energy engineers and companies are constantly surprised by people’s reactions to their plans and ambitions.

To revamp energy models to include people is a massive computational undertaking, even if we knew how to redesign the algorithms or how to effectively engage diverse stakeholders in energy decisions. But we don’t.

Again, maybe AI can help.

There will be other challenges. The data in question are proprietary and batched up by company, by system, by sector, and by country.

Maybe AI can help.

We’ll need to bring in experts from an enormous array of disciplinary and corporate backgrounds and capabilities, synthesize their insights, track data and build models that cut across complex, interdependent systems, from energy, food, and water to transport, manufacturing, and AI itself, and then feed the results back out to networks of decision-makers across all of these industries.

Maybe AI can help.

We’ll need ways of helping broad publics, everywhere, understand the challenges and the trade-offs of reshaping local and global technological systems, imagine the possibilities of building different kinds of futures, collaboratively make decisions, and rethink, redesign, and adapt when problems arise.

Maybe AI can help.

We’ll need training data that don’t automatically assume that sustainability is about human values or natural ecosystems—but rather about the technological webs that we’ve woven between them.

Maybe AI can help.

I’m sure Google and the Stockholm Resilience Center mean well. And I’m sure AI does, in fact, have lots to help us learn about nature.

But nature’s not the problem. We are. Or, more precisely, our relationships with technology.

Maybe AI can help.

We’ll only know if we point AI away from the Earth and, instead, toward the technosphere.

It’s that other engineered planet—the one that humanity has built out of our individual and collective relationships with technology—that we need to learn more about here: how it works, how it might be redesigned to work more sustainably in the future, and how to get our techno-human selves from here to there.

And maybe, just maybe, this is where AI can help us reach beyond our limitations to enable us to build a future beyond our expectations.

---

Clark A. Miller is a theorist and designer of techno-human futures. From 2020-23, he served on the US National Academies Committee on Accelerating Decarbonization of the US Energy System. He is professor and director of the Center for Energy & Society in the School for the Future of Innovation in Society at Arizona State University.