Should we let AI take our jobs…if it generates less carbon?
Fixing Carbon Newsletter

Should we let AI take our jobs…if it generates less carbon?

Some researchers believe replacing human workers with AI has a climate upside. Others think the low-carbon solution is to hit the off switch.
January 14, 2024

Sign up for the Fixing Carbon, a twice-a-month newsletter here

Most jobs have a hefty climate impact. By the time you add up your commute, your computer, heating and lighting, your employment may be responsible for more emissions than your personal life. Now some research suggests that AI systems could replicate some work without the carbon overhead. A company striving for net zero could be tempted to dump humans and embrace technology instead. But are the numbers reliable? And what might be the carbon footprint of all those newly liberated workers?

 

• • •

AI can bring some surprising efficiencies

1.  Thousands of times less carbon. Researchers at the University of California Irvine compared the carbon emissions of human writers and artists beavering away on their laptops, to the latest generative AI systems producing words and images. They found that the Midjourney AI emits approximately 2900 times less carbon dioxide than a US artist to produce a finished illustration, and 370 times less than one based in India. The calculation is based on the time for a person to draw an image and their average daily carbon emissions—thus the difference around the world. When it comes to writing, the same is true: a US author has over one thousand times the climate impact of ChatGPT. “Even relying on cautious assumptions, humans produce far more emissions when engaging in some of the same tasks,” wrote lead author Bill Tomlinson.

2.  Robots are greener than humans. There can be a similar effect when robots replace humans in manufacturing and physical work. AMP Robotics has deployed over 300 AI-guided robots to identify and recover recyclables, reports CBS News. The company says that its robots collect materials up to twice as fast as humans, helping to avoid nearly 1.8 million metric tons of greenhouse gas emissions. A large study in China in 2022 found that the use of both industrial robots and AI have a negative effect on carbon intensity and that the “the labor substitution effect of AI is stronger for labor-intensive industries.” Whether you’re a knowledge worker or on the assembly line, robots might do your work faster and with fewer emissions.

3.  Peak humanity is already here—it’s just unevenly distributed. The population in many developed countries has already peaked, with the whole planet now due to reach its largest population in 2086 rather than 2100. The reality is that the West already has an aging population, with fewer working-age humans—and the rest of the planet is catching up fast. When these demographics start to bite, we might need AI workers to not only decrease our emissions but maintain our quality of life.

 

  • • •

AI is just another gadget—and gadgets need power

1.  You can’t turn humans off when they’re not working. The problem with like-for-like analyses of AI and humans is that they’re not alike. Every human replaced by an AI system is still eating, moving, and living their life, with all the attendant emissions. While there are some big carbon savings from not having to commute (which we do less of these days anyway), the overall climate impact will depend on whether laid-off workers live it up, or sit quietly at home.

2.  AI has a staggering carbon footprint. Regardless how efficient an AI is compared to a human employee, most AI emissions are enormous and still increasing. Training the Bloom large language model over five months in 2022 took as much energy as powering a US home for over 40 years—and the most extensive AI training runs have been doubling the amount of computing power they use about every 3.4 months, reports MIT Sloan Management Review. Every few dozen conversations with ChatGPT “drinks” half a liter of water at data centers, reports the Wall Street Journal, and generating one AI image takes about the same energy as fully charging your smartphone.

3.  We already have enough brains on the planet. With nearly 8 billion intelligent, adaptable and motivated humans on Earth, we don’t really need any more processing power. What if we put our energies into boosting prosperity and helping more people reach their full potential, instead of chasing another technological dream? 

 

• • •

What to keep an eye on

1.  Brain-like hardware. A new class of computer chips inspired by the structure of the human brain promises to reduce AI’s growing carbon footprint. So-called neuromorphic processors can be dozens of times more energy efficient (and tens of times faster) than today’s silicon.  

2.  UBI. Some tech billionaires, including OpenAI founder Sam Altman, believe that AI will usher in a golden age of leisure for humans, and that some form of Universal Basic Income—free cash payouts— will be necessary for social equity. Numerous UBI pilot studies are already underway in the US and abroad.

3.  AI unlocking carbon savings. AI is about far more than chatbots and plagiarized images. There’s a global AI for Climate initiative—and plenty of real-world examples. A collaboration between Google and American Airlines used AI-based predictions to reduce heat-trapping contrails from 70 jet test flights by over 50%, AI can balance electrical grids to minimize fossil fuel use, and even unlock a wild and sustainable plant-based diet

Top image: ©Anthropocene Magazine

What to Read Next

Anthropocene Magazine Logo

Get the latest sustainability science delivered to your inbox every week

Newsletters

You have successfully signed up

Share This Article