🗃️ You’re viewing an archived issue of our newsletter, The Pipette. View all issues.
Environmental Impact of AI
👋 This is The Pipette – a monthly newsletter for product builders with links so good their ideas warrant a reply, a forward, or even a discussion in real life.
This isn’t necessarily an AI-focused newsletter, but it’s been a hard topic to ignore. Generative AI is starting to touch every industry to some degree, and there’s a lot to cover. Today we’re going to hone in on the environmental impact of AI.
With all the buzz around AI, urging us to either jump on the bandwagon or approach with caution, we’re excited to share our perspective and “hot takes”, which you can think of as conversational snippets we’d say in person (how vulnerably un-internet-like is that?).
Keep in mind, our views are open to change, and we’re eager to hear your thoughts – simply reply to this newsletter with your own hot takes or other feedback you may have!
AI needs energy, and lots of it
It takes a lot of compute power to train and run these AI applications, especially with how popular they have become (ChatGPT reached 1 billion monthly visitors in Feb. 2023 – source). Researchers are starting to quantify exactly how much energy is needed, and the results are mind-boggling:
- ChatGPT drinks a water bottle every 5-50 prompts – a look into OpenAI’s water usage at Microsoft’s US data centers shows how many resources are needed to run ChatGPT. Microsoft reported a 34% spike in water usage from 2021-2022, likely tied to AI research. Google reported a 20% increase in the same period.
- Making an image with generative AI uses enough energy to fully charge your phone. That old saying about a photo being worth a thousand words holds up here too; creating 1,000 text times would only use 16% of a phone charge in terms of energy. The article also covers an upward trend in energy usage with these tools. And with the release of OpenAI’s Sora text-to-video generator, it will likely get worse.
Hot take: As capabilities and demand for generative AI grows, so will its environmental impact. Public pressure, along with the desire to maximize compute resources, will force companies to find ways to use less energy. End users can also help by writing more efficient prompts!
Other Context:
- We cover the broader impacts of generative AI adoption on our site. The article goes into the cost of it, prompt engineering, and how businesses are reprioritizing to adapt.
- While looking for images about AI, we found a great collection of art commissioned by Google’s DeepMind. The goal is to have artists depict their representation of AI. You won’t find any robots on it!