The issue with finding that number, as we explain in our piece published in May, was that AI corporations are the one ones who’ve it. We pestered Google, OpenAI, and Microsoft, but each company refused to supply its figure. Researchers we spoke to who study AI’s impact on energy grids compared it to attempting to measure the fuel efficiency of a automobile without ever with the ability to drive it, making guesses based on rumors of its engine size and what it appears like taking place the highway.
MIT Technology Review
But then this summer, after we published, a wierd thing began to occur. In June, OpenAI’s Sam Altman wrote that a median ChatGPT query uses 0.34 watt-hours of energy. In July, the French AI startup Mistral didn’t publish a number directly but released an estimate of the emissions generated. In August, Google revealed that answering a matter to Gemini uses about 0.24 watt-hours of energy. The figures from Google and OpenAI were just like what Casey and I estimated for medium-size AI models.
So with this newfound transparency, is our job complete? Did we finally harpoon our white whale, and if that’s the case, what happens next for people studying the climate impact of AI? I reached out to a few of our old sources, and a few latest ones, to seek out out.
The numbers are vague and chat-only
The very first thing they told me is that there’s so much missing from the figures tech corporations published this summer.
OpenAI’s number, for instance, didn’t appear in an in depth technical paper but somewhat in a blog post by Altman that leaves plenty of unanswered questions, reminiscent of which model he was referring to, how the energy use was measured, and the way much it varies. Google’s figure, as Crownhart points out, refers back to the median amount of energy per query, which doesn’t give us a way of the more energy-demanding Gemini responses, like when it uses a reasoning model to “think” through a tough problem or generates a very long response.
The numbers also refer only to interactions with chatbots, not the opposite ways that individuals have gotten increasingly reliant on generative AI.
