Developing and maintaining AI-based conversational beings ̶ such as ChatGPT ̶ will significantly increase global energy demand. In the interests of global sustainability, that additional power must be from renewable sources. Original graphic (Monica Whipple and David Turner). Image Credits: Circuitry, Wind Farm, Solar Panels, Pylons.
When the sheriff character in the original “Jaws” movie first sees the giant shark, he exclaims to the captain “You’re gonna need a bigger boat”.
An analogous statement regarding the energy requirements associated with the coming proliferation of conversational virtual beings (based on Artificial Intelligence) is that the technosphere is going to need a bigger power supply.
By virtual beings I mean all the digital, language-capable, denizens of the emerging metaverse (broadly defined), including chatbots (like ChatGPT), AI-assisted search engines (like Perplexity AI), and AI-based residents of Meta’s visor-enable virtual reality world. Coming down the line are speaking holograms, and holodecks (as in Star Trek).
The process by which these advanced digital creatures learn to speak is based on development of neural networks that are trained with a large body of textural information (like Wikipedia, books, and an array of content available on the Internet). Training means determining statistical relationships between the occurrence of different words in the training text, which the algorithm then uses to formulate a response based on keyword inputs (queries).
Training a large language model such as ChatGPT requires a hefty input of computing power because it involves extensive trial and error testing. Chatbots affiliated with AI-assisted Internet searches use not just a pre-trained language model but also integrate the search output into their responses. This kind of processing will be energy demanding (perhaps 5 times greater than for a standard search), which will add up considering the billions of searches made per day.
If these virtual beings were only going to be used by a minority of people (such as now visit Meta’s colony in the metaverse), the power draw would be minor. But, very likely, their seductive appeal will be so great (albeit with an occasional hint of menace) that they will become a standard feature of ordinary life. Just in the field of education, there is vast potential for inspiring and informing students using dialogic Chatbots.
The overshoot model argues that global energy consumption should be reduced rather than expanded because of the many negative environmental externalities (unaccounted for damages) caused by energy production ̶ from both fossil fuel and renewable sources.
However, at least for electricity, that seems unlikely given the burgeoning energy demand in the developed world noted here, and the aspiration to raise standards of living in the developing world.
Since 66% of global electricity production is still based on combustion on fossil fuels, any increase in electricity consumption will tend to result in more greenhouse gas emissions and more societal problems with climate change. The obvious conclusion in that new energy demand must be met by nonfossil fuel sources like hydro, wind, solar, geothermal, and nuclear fission. Companies such as Google, Microsoft, and Meta that are building the metaverse will experience huge increases in energy consumption in the near future; they should be held to their commitments to run on carbon neutral power sources.
New energy technologies that could contribute to a clean global power supply in the coming decades include geologic hydrogen and solar energy from space. These sources, however, will require long-term investments in research and development.
The global renewable energy revolution is off to a good start and has a bright future, but it will require steady political pressure to 1) stop building new fossil fuel burning facilities, 2) replace aging fossil-fuel-based infrastructure with renewable sources, and 3) build new renewable energy sources that can accommodate the increasing demand that is surely coming.
National governments the world over have made political commitments to reduce greenhouse gas emissions significantly in the next few decades. Because the generation of electricity, i.e. the power sector, is currently one of the largest anthropogenic sources of CO2 emissions (due to its reliance on coal and natural gas burning power plants), a great deal of research and investment is directed towards power sector decarbonization.
There are many pieces to the technical puzzle of how to decarbonize the power sector, and the optimum answer will differ depending on location and available resources. But generating electricity while avoiding fossil fuels altogether is entirely feasible.
In that regard, I was happy to see news of a funded power project that nicely weaves together many of the critical components needed to deliver carbon-free electricity at grid scale (Figure 1).
The facility in this case is being built in French Guiana by a consortium of private firms. The exciting thing to see is the co-location and integration of five key power generation components: (1) an array of solar panels, (2) an electrolyzer to produce hydrogen, (3) a hydrogen gas storage capability, (4) a hydrogen fuel cell that generates electricity, and (5) a short-term battery energy storage system. Functioning together, these components will provide a 24/7 baseload supply of carbon free electricity (10,000 households worth).
The solar array collects sunlight. Most of the energy is fed into the local electricity grid, but a portion is directed to the electrolyzer to split water molecules into oxygen and hydrogen. The hydrogen gas is stored on site. At night, the hydrogen is supplied to the fuel cell generator. The short-term battery storage system helps maintain a steady flow of energy as needed.
This kind of facility largely solves the intermittency problem for renewable solar energy. Its design could be adapted to other renewable energy sources with an intermittency problem, notably wind energy farms. Excess hydrogen could potentially be transported to other locations by pipeline or in liquid form.
Successful operation of the facility (slated to open in 2024) will provide a model that potentially could be scaled up and widely adopted. Since garnering the political will and financing for renewable energy development is still a significant challenge, the completion and operation of this power plant would send of strong signal about the feasibility of decarbonization to government, industry, and sources of investment.
News that this facility is actually under construction inspires the feeling that the global we (such as it is) can indeed accomplish a needed renewable energy revolution.
When I was 20 years old, I picked up a paperback version of
“Life and Energy” by Isaac Asimov. This
lucid scientific description of the chemical basis for life was very
compelling, indeed, it helped inspire me to pursue a career in biology and
ecology. The time around its publication
in the mid 1960s was quite exciting in biology because the fields of
biochemistry and cell biology were in full flower; scientists had worked out
the role of DNA in regulating cellular metabolism and had achieved a good
understanding of the chemistry of photosynthesis and respiration.
Metabolism is broadly defined as the chemical machinery of
life, the networked sequences of chemical reactions that build and maintain living
matter. Biologists think of living
matter in terms of levels of organization – from cells, through
organisms, communities, and the biosphere.
In Asimov’s days, the concept of metabolism was mostly applied at the level of the cell or organism. However, ecologists in recent years have also applied it in the context of ecosystems and the Earth system as a whole. Here I would like to consider metabolism at the scale of the technosphere.
An ecosystem is a biogeochemical cycling entity, e.g. a pond
or a patch of forest. Like an organism,
it requires a source of energy and it cycles nutrients such as nitrogen from
one chemical form to another. Strictly
speaking, it is the biota (the set of all organisms) that in a sense has a
metabolism. Component organisms are
classified into nutrient cycling guilds — most simply as producers (photosynthesizes),
consumers, and decomposers.
Ecosystem metabolism can be described in terms of energy fluxes, as well as the stocks and fluxes of key chemical elements. The element carbon plays a central role in ecosystem metabolism. Its cycle extends from the atmosphere, through plants, to animals, and to decomposers, then back to the atmosphere.
The ecosystem carbon cycle. Image credit, Figure 4.1, The Green Marble, David Turner, 2018, Columbia University Press.
At the scale of the Earth system, we can likewise talk about
biogeochemical
cycling guilds and the associated biogeochemical cycles. Biosphere metabolism is based on
photosynthesis on the land and in the ocean.
Biosphere driven element fluxes help regulate the atmospheric chemistry,
ocean chemistry, and global climate.
Despite repeated intervals in the geologic record of Hothouse
Earth and Icehouse Earth, the metabolism of the biosphere has run steadily for
over 3 billion years.
Quite recently in geologic time, a new sphere has emerged within
the Earth system. This “technosphere” is
the cloak of technological devices and associated human constructs that has
come to cover the Earth. Like the biosphere,
it
has a metabolism.
We can think about technosphere metabolism in terms of three
key factors: energy flows, materials cycling, and information processing.
Humans are a part of the technosphere and mostly benefit
from its metabolism. The technosphere
produces a vast array of goods and services that support billions of
people. However, the metabolism of the technosphere
has begun to disrupt the formerly background global biogeochemical cycles. It is effectively now a geological force and
the changes it has precipitated are not necessarily favorable to advanced
technological civilization. Notably, the
delivery of vast amounts of CO2 into the atmosphere is destabilizing
the global climate. A course correction in
the evolution of the technosphere is required.
Energy flow into the technosphere is predominantly from the
combustion of fossil fuels (coal, oil, and natural gas). The problem is that the resulting source of
carbon to the atmosphere is orders of magnitude greater than the background
source from volcanoes. The background
geologic sink for CO2 by way of mineral formation in the ocean
depths is likewise small.
Some of the technosphere-generated CO2 is sequestered
by land plants and in the ocean, but most of
it is accumulating in the atmosphere and causing the planet to rapidly warm. The technosphere has begun to disrupt the
entire Earth system.
The solution, as is well known, is conversion of the global energy infrastructure to renewable energy sources (solar, wind, hydro, geothermal, biomass, and renewable natural gas). That conversion is a daunting task but technically it can be accomplished. The challenge is as much to economists and politicians as it is to engineers.
Converting from fossil fuel-based energy sources to renewable energy sources. Image credit for power plant and wind turbines.
The materials cycling factor in technosphere metabolism is problematic because the technosphere as currently configured is not effective at recycling its components. Unlike the biosphere, in which nutrients are cycled, there is often a one-way flow of key chemical elements in the technosphere − from a mineral phase, to a manufactured product, to a landfill.
The problem is that sources of the technosphere components
are not infinite. Building the next mega-mine
to extract aluminum degrades the biosphere, a key component of the global life
support system.
Again, this is a largely solvable problem using advanced
industrial practices. We now speak of
the emerging circular
economy and of dematerializing
the technosphere. More comprehensive
recycling may require more energy than dumping something into a landfill, but
the potential for renewable energy sources is large.
The information processing aspect of technosphere metabolism
refers to its regulatory framework. Regulation
requires information flow, a receiver of that information, and a mechanism to
act on it.
Homeostasis at the level of an organism is a clear case of
regulation. Homeostasis of internal
chemistry, such as the blood sugar level in mammals, depends on factors
including signals based on chemical concentrations, DNA-based algorithms to
formulate a response, and organs that implement a response.
Ecosystems also self-regulate in a sense. Disturbances (e.g. a forest fire) are
followed by vigorous regrowth. As a
result, nutrients that are released in the process of the disturbance are
captured and prevented from loss by leaching.
Damaged ecosystems, say that lack species specialized for the early
successional environment, may deteriorate after a disturbance.
At the global scale, Earth system scientists have long
debated the issue of planetary homeostasis.
James Lovelock famously hypothesized that indeed the Earth
system (Gaia) is homeostatic with respect to conditions that favor
life. His idea inspired much research,
and many significant biophysical feedbacks to global change have been
identified. The biosphere clearly exerts
a strong influence on global climate by way of its impacts on greenhouse gas concentrations,
specifically through its role as an amplifier
in the rock
weathering thermostat.
A new research question concerns the degree to which the technosphere
is homeostatic. Continued
exponential growth in many of the indices of technosphere metabolism is
suggestive of inadequate regulation. To
begin with, the regulatory capability of the technosphere is obviously diffuse
and underdeveloped.
Monitoring is a necessary component of effective management
systems but the self-monitoring capability of the technosphere barely existed
until quite recently. An anomalous
growth in the atmospheric CO2 concentration was measured in the late
1950s by atmospheric chemist Charles David Keeling. This observation was the first clear signal
of technosphere impact on the Earth system.
The Landsat series of
satellite-borne sensors that monitor land cover change, e.g. deforestation
and urbanization, was initially launched in 1972. These satellites have since tracked the explosive
spatial expansion of the technosphere. A
fleet of other satellites now monitors many other features of the global
environment.
From synthesis efforts by agencies such as the United Nations
Food and Agriculture Organization, we have good observational data on the
growth of key technosphere variables like global population size and energy use.
As far as a decision-making organ for the technosphere, one
barely exists at present. You might say
that market-based capitalism is the organizing principle of the current technosphere. Everyone wants cheap, plentiful energy (the
demand side) and the global fossil fuel industry has managed to keep ramping up
the supply. Under the current neoliberal
economic regime, the environmental costs are externalized, and no global oversight
is imposed.
However, a new constraint has arisen. The scientific community has built Earth
system models to refine our understanding of Earth’s biophysical regulatory
mechanisms and to simulate effects of various greenhouse gas concentration scenarios. These simulations make clear that uncontrolled
emissions of CO2 from fossil fuel combustion must cease or advanced technological
civilization will be imperiled. The 2015
Paris Agreement on Climate Change was a step towards reining in the technosphere,
but the influence of that international agreement is not commensurate with the
challenges of current global environmental change.
An essential feature of a needed paradigm shift regarding technosphere
regulation is the development of a global environmental governance
infrastructure. The technosphere is
having global scale impacts on the environment and must correspondingly be
evaluated and regulated at the global scale.
A proposed World
Environment Organization would not necessarily supersede the traditional nation-state-based
architecture of global governance, but it could go a long way towards the
required scale of integration needed to address global environmental change
issues.
A revamped technosphere metabolism must be built over the
course of the 21st Century in which the energy sources are
renewable, the material flows are cyclic, and the regulatory framework is rooted
in an understanding of limits. Societies
are more likely to change under extreme circumstances, and the economic shock
of the 2020 coronavirus pandemic will certainly qualify as extreme. As the global economy recovers, there will be
significant opportunities to change technosphere metabolism. Let’s hope they are not wasted.
