AI is here, but let’s talk about whether we should really be using it.
Summary: This article explores the environmental impact of AI, weighing its time-saving and productivity-boosting benefits against its carbon footprint.
I have mixed feelings about using and promoting the use of AI. It’s an incredible tool for supporting work —both day to day, and in climate response projects like charting methane emissions, or identifying buildings damaged in natural disasters— but it comes with a concerning environmental impact (not to mention bias, privacy and misinformation issues, worthy of a whole other deep-dive).
What’s the environmental issue?
There are a few major problems here:
- Energy consumption: training a single AI model can emit as much carbon as five cars in their lifetimes
- Water usage: data centers need an abundance of water for constant cooling (eg. ChatGPT needs an estimated 500ml of water for every 20-50 questions asked)
- Electronic waste: AI hardware needs replacing often which contributes to e-waste (an estimated 57 million tons are generated each year, about the weight of the Great Wall of China)
- Environmental inequity: the environmental impact of AI can disproportionately affect certain communities
How can we reduce these impacts?
There are certainly things the key players can do:
- Renewable energy: data centers can be powered using renewable energy sources
- More efficient AI: algorithms can be improved to use less power
- Carbon offsetting: companies can report their emissions and invest in carbon offsets
For most of us, these things are beyond our control so there’s an ethical choice to make around whether we use this technology or not. Since the technology does exist (and appears to be here to stay), should we harness its power for personal innovation and support at home or work? Or boycott these tools?
For now, I’m acting in the hope that there is a pathway for sustainable AI through projects like Expeditions in Computing (which aims to reduce the carbon footprint of computing by 45% within the next decade) and the take up of the UNEP’s recommendations around impact reporting, regulation, efficiency, offsetting emissions, and connecting AI and environmental policy.
Ever the optimist. But I feel like I have to be. What’s your stance?