Will AI Bring Plentitude Or Further Imperil The Planet?

So far, the leaps in energy-intensive computing power far outpace the lagging shift to a green economy.

Nusha Ashjaee for Noema Magazine
Credits

Nathan Gardels is the editor-in-chief of Noema Magazine.

Frugality — the prudent husbanding of resources to meet needs — has long been framed by ecologists as the only realistic alternative to the endless wants of consumer capitalism that are belching ever more warming emissions into a biosphere on the brink of breakdown. 

But what if a kind of frugal plentitude, a clean and equitable condition of post-scarcity, could be achieved through the awesome capacity of an AI-driven internet of things to drastically reduce the cost of producing goods, eliminate waste by precisely matching supply with demand and diminish the outsized carbon footprint of the industrial age through smart factories and cities wired up with sensors for energy efficiency?

In Noema this week, we ponder this “dream of plentitude” imagined in a tale by Chinese science fiction writer Chen Qiufan. The excerpt is drawn from his new book, “AI 2041,” written with AI guru Kai-Fu Lee, with whom we also spoke recently.

In his analysis that follows Chen’s story in their book, Lee projects forward to the middle of the century:  “The age of plenitude will arrive when most things are no longer scarce, can be produced for next to nothing, and — most important — are made available freely or cheaply to all people. These ‘almost free’ things will begin with necessities like food, water, clothing, shelter, and energy. Over time, plenitude is a process in which more and more goods and services are provided to more and more people, as technologies advance and costs come down to grant new free ‘indulgences’ every year. I expect that plenitude would start with these necessities and gradually expand to provide a comfortable and gracious lifestyle for all, encompassing a provision for transportation, clothing, communication, healthcare, information, education and entertainment.”

In Chen’s fictional tale, the Australian government launches a project called Jukurrpa to give citizens all of these benefits for free through a social currency system that meets physical needs with a “Basic Life Card.” Beyond that, another denomination called Moola rewarded citizens for voluntary work and encouraged them to “lead purposeful lives filled with love and belonging, while becoming more empathetic and compassionate.” 

Lee argues that this vision of post-scarcity is not so far-fetched as it sounds, citing science fiction author William Gibson’s observation that “the future is already here — it is just not very evenly distributed.”

“In 2020, the United States discarded $218 billion worth of food, while the cost to eliminate hunger in the U.S. has been estimated at just $25 billion per year,” Lee notes. “In the United States, there are more than five times as many unoccupied houses as there are homeless people. So we already have theoretical plenitude in 2021 for food and shelter in the United States.” Thus, once AI computation can ferret out the waste of inefficient allocation by creating a frictionless pairing of need and supply, we can arrive at a much fairer society. 

Leaving aside the considerable social and political blockages to this vision of an evenly distributed future, my conversation with Lee focused on the myth of “dematerialization” — that the digital economy ahead will be virtually free of real-world externalities, cleaner and less resource-intensive than the dirty and rapacious era of mining and manufacturing. 

Digital downloads of all manner of services to our smartphones appear to be “almost free.” We can carry our devices lightly in our hands. They have the computing power that once required a building-sized mainframe at a tiny fraction of the cost. But now there are billions of them across the planet, crammed with materials that must be mined from the Earth, manufactured in polluting factories and shipped thousands of miles to consumers who replace them with newer models in short order. Thus, “less” is far more resource-intensive the more ubiquitous the smaller devices become. And to run all our searches, no less the intensive computations of AI, vast amounts of energy are devoured by server farms where the data is stored, mostly powered by fossil fuels. 

“I agree with this concern,” says Lee. “If we want to continue to develop advanced forms of AI, computation capacity must grow rapidly. Just to train an AI algorithm or deep learning model can cost millions of dollars and put a lot of stress on server farms.”

He acknowledges that “we’re a little stuck right now because the state of the art requires so much computation. As ordinary companies seek to make breakthroughs like Microsoft and Google and OpenAI have done, the amount of computation power needed will go up. It is therefore important to invest in efficient software and tools that can make top-end technologies more practical, perhaps slightly reducing performance while still providing the computation they need.” The chief challenge, as Lee sees it, is how to make AI smarter without just throwing more data and computing power at it. His hope rests on the iterative tweaking of algorithms that improve the performance of AI at a “geometric pace.”

Lee also looks optimistically to distributed energy that combines solar with advanced battery technologies beyond lithium. “We might reduce energy costs to 10% of what they are today over the next few decades,” he says. “Even that may not be enough, but we have to balance the two.”

Here, the faltering footsteps of offline reality intrude. While computing power grows in exponential leaps, the pace of transition to a green economy that would cleanly fuel that growth lags far behind. Nothing but the availability of data and the arrangements of ones and zeros or quantum calculations stands in the way of AI programming. By contrast, political gridlock driven by deeply entrenched vested interests, not to speak of the lifestyle inertia of consumer society, frustrate any rapid advance toward a green future. 

The present global supply-chain bottleneck spotlights the highly uneven integration of the frictionless digital economy with the real physical world of constraints on the production, transportation and distribution of goods. It is both a metaphor and premonition of the near future in which the acceleration of the resource-sucking data economy outpaces the energy transition even as we approach the tipping point of climate change.

In the end, there is no escaping the hard reality that AI-assisted post-scarcity can only be reached if renewable energy advances in tandem with computing power. If not, intelligent machines will end up worsening a planet in peril instead of paving a path to plentitude.