In this episode I dig into the environmental impact of using ChatGPT: energy use, water consumption, and carbon emissions. I walk through the research, acknowledge the risk of whataboutism, and explain why comparisons to other systems matter when we're talking about true impact. Adopting a both/and approach, this episode is absolutely not about letting ChatGPT off the hook, it’s about putting things in perspective so we can take action on what will actually move the needle.
Main Topics Covered
Why this is episode 2 and why the topic mattersRamit Sethi’s $30K vs $3-question analogyThe math behind ChatGPT’s energy useThe origin (and problems) with the “3Wh per query” statWhat .3Wh per query actually looks like in real-world useThe difference between inference and training impactHow ChatGPT kicked off the AI arms raceWhere most AI energy actually goes (spoiler, it's not ChatGPT)The hidden energy behind everything else we useHow water usage is measured across three scopesChatGPT’s water use vs typical U.S. electricity and food systemsWhy carbon emissions highlight the need for more transparencyThe repeating citations problem in media coverageWhat actually makes the biggest environmental differenceConcerns about future energy use from agentic modelsWhy transparency from companies mattersThis episode as a call for participatory awareness, not panicLinks & Resources For This Episode
Check out the full environmental impact resources list on my website Subscribe to the ChatGPT Curious NewsletterSubmit a QuestionVisit the WebsiteFeeling curious AND generous? Click here to support the podcast.