Member-only story

The Ecological Footprint of Outsourcing Creativity: AI’s Hidden Environmental Costs

Running, Training, and Using LLMs Cost Concerning Amounts of Power In an Already Strained System

--

Photo by Dominik Scythe on Unsplash

The United States is running out of electricity. In March, The Washington Post published an article expressing concern about the surge in power use in several states that are struggling to keep up with demands.

Our power grids were already due for an upgrade, with most of them constructed near the end of the 20th century and a reported 70% of transmission lines approaching the end of their lifecycle. The effects of our aging system are strikingly obvious to those who’ve lived through them. Power outages are far more common now than 25 years ago, with an estimated 64% increase in the US since the 2000s.

And, to add salt to the untreated, festering wound that is our energy infrastructure, we vastly underestimated how much power we’d need in the 2020s. We didn’t expect the rapid rise and proliferation of artificial intelligence, and certainly didn’t consider the massive amounts of power it would require. Those costs in question are staggering already and, are now, expected to rise astronomically in the next decade as every company rushes to include AI in their pipeline and competitors rise to develop their models amongst the Artificial Intelligence Gold Rush.

As we enter the AI era, we must consider the electrical toll such technology can have on our already straining power grid. Even more so, it’s imperative that we consider what impact that might have on our already straining environment. We must consider the ecological footprint of creating and using LLMs.

Deep Learning and The Cost of Running an Artificial Mind

Every time you access a website on the internet, you’re directing your computer to go to an address on a server somewhere out in the world and bring back that information. Typing a URL into the address bar tells your browser of choice where you want to pull content from. It then takes that provided direction, grabs that little bit of digital information, then delivers and displays it on your screen as fast as the information will allow.

--

--

Olivia Louise Dobbs
Olivia Louise Dobbs

Written by Olivia Louise Dobbs

Naturalist who writes about STEM. Curriculum developer, Biostats graduate student, author, general purpose nerd. 🦜New blog every other Friday!

No responses yet

Write a response