Science 3 min read

A Single AI's Carbon Emission is Nearly 5x Greater Than a Car

Andrey_Popov / Shutterstock.com

Andrey_Popov / Shutterstock.com

Carbon emission isn’t the first thing that comes to mind when discussing artificial intelligence. It’s algorithms after all, isn’t it?

But these algorithms power software that run on electricity, and when it comes to AI programs, their energy needs extend to even before they’re up and running.

AI energy consumption levels become more concerning when we look at the broad spectrum of industries that are betting on AI to conquer new scientific grounds.

Deep neural networks are good at mining data, but can we afford the environmental cost?

Energy-Intensive Deep Learning and Carbon Emission

To create efficient deep neural networks, you need to train them on large datasets of whatever you need them to train on. For that, you need enormous computational resources that require similarly large amounts of energy.

And this is where things get, well, polluted.

Researchers at the University of Massachusetts, Amherst, found that training AI models emit nearly as much carbon in the air as five cars, and that includes fuels and the cars’ manufacture process itself.

The team trained four different AI models for one day each, took sample measurement, and extrapolated to assess their total energy consumption over a life cycle, taking in regard the total training time as reported by AI developers.

To estimate each AI model’s carbon footprint, they referred to the average carbon emissions of electricity production in the United States.

They came down to a startling figure, finding that an average, off-the-shelf deep learning software can emit over 626,000 pounds of carbon dioxide.

That’s a massive volume of carbon dioxide in the air that makes dirty tailpipes of gasoline and diesel-powered cars suddenly look more climate-friendly.

AI Needs Renewable Energy

UMass Amherst’s Emma Strubell is one of the research team and co-author of the paper (preprint) that they’ll present at the annual Association for Computational Linguistics conference in Florence, Italy. Strubell said:

“I’m not against energy use in the name of advancing science, obviously, but I think we could do better in terms of considering the trade off between required energy and resulting model improvement.”

Researchers pinpointed that the most energy-intensive of all AI training processes is Neural Architecture Search (NAS). It appears that while AI models in NAS perform specific tasks, they require the most energy. This type of processes is used mainly by big tech firms that can afford the huge power bill of their AI models.

This includes, for example, Amazon’s voice assistant Alexa and Google’s translator and their other AI algorithms and platforms.

So the question isn’t really about how much energy AI consumes, and the subsequent carbon emission, but the source of this energy.

How green is Amazon’s and Google’s energy mix overall?

Well, it’s as green as the U.S. energy mix. But, there’s some substantial change in this regard as these companies are taking initiatives to use more renewable energy and cut CO2 emissions of their data centers.

Amazon, which is investing in solar and wind farms, says it has “exceeded 50% renewable energy usage for 2018, and reiterated its commitment for reaching 100% renewables.

Pursuing its goal of running on 100% renewable energy, Google also is greening its grid.

That all sounds good until you know that 100% renewable energy doesn’t guarantee 100% zero-carbon energy, but that’s another story.

Read More: Game Changer: Renewable Power Sources Become Cheapest Global Energy Source

First AI Web Content Optimization Platform Just for Writers

Found this article interesting?

Let Zayan Guedim know how much you appreciate this article by clicking the heart icon and by sharing this article on social media.


Profile Image

Zayan Guedim

Trilingual poet, investigative journalist, and novelist. Zed loves tackling the big existential questions and all-things quantum.

Comments (0)
Most Recent most recent
You
share Scroll to top

Link Copied Successfully

Sign in

Sign in to access your personalized homepage, follow authors and topics you love, and clap for stories that matter to you.

Sign in with Google Sign in with Facebook

By using our site you agree to our privacy policy.