by Mark Douglas
Ask
Gemini to create an image of the word “ecosystem”
and it is likely to produce a still-life: wild animals in a bucolic setting of
rocks, trees, and water beneath blue skies. Ask Gemini to create an image of
the phrase “AI ecosystem” and it is likely to produce
an image of the “cloud”: electrons whizzing
through and between circuits of data chips with a ghostly image of a humanoid
robot face or hand benevolently and all-but-passively guiding them.
Both
images are lies. One problem is that static images cannot convey dynamic
processes—and both ecosystems and AI are dynamic. Another is that Gemini, unsurprisingly,
favors some organisms (the so-called “charismatic megafauna”)
and excludes others. Most of the organisms that make life in an ecosystem
possible are invisible: hidden, microscopic, or offstage. Then there is a final
falsehood promoted by most images of ecosystems: the exclusion of human beings
and their effects from those images. But there are simply no parts of the world
that human beings and their stuff have not touched. Any image of any part of any ecosystem on
Earth that doesn’t include humans and their impacts is a
lie.
The
same lies that afflict images of natural ecosystems beset images of AI
ecosystems. Such images fail to reproduce the dynamism of AI systems and the
myriad forces and energies at work powering AI and its products. They favor
some ways of seeing and things to see over others. They ignore the fact that
there are politics to AI, determined as its functions are by algorithms that
are neither value-neutral nor purposeless. And even where those images display
something human-looking, they replace the uniqueness of human faces and bodies
with reproducible and simplified facsimiles of such things.
Those
reproducible and simplified facsimiles point towards a problem with many
contemporary discussions of AI: they treat it as ethereal and noncorporeal. But
AI ecosystems are not ethereal and noncorporeal.
AI
is neither disembodied nor separable from its impacts on bodies. The lithium
used for the batteries necessary to use AI is mined around the world, including
from places like Congo, Mongolia, and Indonesia. The rare earth metals
(lanthanum, cerium, terbium, yttrium, etc.) used in the creation of processers
and circuits of computers as well as componentry of modern communication
systems like fiber-optic cables, communications towers, and satellites, are
mined widely but often in politically fragile systems. AI is composed not
simply of algorithms, but of elements--and getting those elements from inside
the earth into computers took countless human beings (many of them working in
oppressive conditions) and extraordinary amounts of energy (mostly produced from
burning fossil fuels), leaving behind vast scars on the earth (despoiling the
natural ecosystems in those areas).
Further,
AI systems require huge volumes of coolant—mostly water—to keep those systems,
especially their data centers, from overheating. Since these data centers
require lots of land and ease of access to infrastructure and because they can
be so noisy, they are often located in deserts and other rural areas: the very
places where water is scarce or needed for other purposes. Water used to cool a
computer is water not being used to irrigate a crop or maintain hygiene—and,
again, moving water, like moving rare earth metals, takes energy.
Then
there is the sheer amount of energy used to power AI. The production, storage,
and dissemination of data takes energy—and AI relies on (and produces) a lot of
data, even compared to other tech systems.
Energy consumption by data centers had doubled by 2023, by mid-2024,
4.4% of all energy used in the U.S. went to data centers and, at current rates
of growth, by 2028, “AI alone could consume as
much electricity annually as 22% of all US households.”[1]
All
these environmental costs are borne not just by the non-human natural world,
but by human beings who are and will be living with their consequences, whether
in the form of polluted communities, politically precarious governance systems,
diminished access to basic human resources like water, or the increasing range
of poor-health consequences due to climate change. Moreover, these consequences
are not borne equally by all persons: the most deleterious effects are borne by
the most vulnerable persons.
All
of which returns to questions about artificial intelligence and ecosystem
thinking. Its very potency, utility, and transformative potential mean AI is
going to be used and its use is going to cause harm to ecosystems. The
inevitability of such use makes moral questions about whether to use AI
unhelpful. We will, increasingly, turn to AI in order to make our lives better
(or at least different).
Instead,
placing our use of AI within ecosystem thinking should prioritize certain types
of concerns and actions:
·
Recognize the harm that our use of AI does
to the nonhuman natural world.
·
Recognize this harm means that we should
include attention to such harm in the ways that we describe and promote it and
as we shape policies and markets around it.
·
Address the environmental costs for using
AI by including obligations towards reparative justice in our use of it.
·
Harness AI to sift through the massive
amounts of data that need to be taken into account when calculating the types
and amounts of harm done to the nonhuman natural world by human beings and by
AI. Indeed, given the volumes of data used in climate modeling, AI could prove
a vital tool in our response to climate change.
·
Seek to lessen AI costs by pursuing
technologies that reduce the environmental load associated with artificial
intelligence.
What
will it look like if and as these and other suggestions are implemented? That
may be hard to say. We are working in strange new territories as we
simultaneously engage powerful and novel technologies and also address a
natural world that is changing in ways unlike any in recorded human history. We
do, though, have a pretty good sense of what will happen if we do not attend to
the impacts of AI on the nonhuman—and also the human—natural world. Those are
costs that even AI would say we should avoid.
Mark Douglas is the J. Erskine
Love Professor of Christian Ethics at Columbia Theological Seminary
[1]
James O’Donnell and Casey Crownhart, “We did the math on AI’s energy footprint.
Here’s the story you haven’t heard.” MIT Technology Review (2025).
Available at https://www.technologyreview.com/2025/05/20/1116327/ai-energy-usage-climate-footprint-big-tech/.
Accessed on Sept. 18, 2025.
No comments:
Post a Comment