Curated from MIT Technology Review — Here’s what matters right now:
Google just announced that a typical query to its Gemini app uses about 0.24 watt-hours of electricity. That’s about the same as running a microwave for one second—something that, to me, feels virtually insignificant. I run the microwave for so many more seconds than that on most days. I was excited to see this report come out, and I welcome more openness from major players in AI about their estimated energy use per query. But I’ve noticed that some folks are taking this number and using it to conclude that we don’t need to worry about AI’s energy demand. That’s not the right takeaway here. Let’s dig into why. 1. This one number doesn’t reflect all queries, and it leaves out cases that likely use much more energy. Google’s new report considers only text queries. Previous analysis, including MIT Technology Review ’s reporting, suggests that generating a photo or video will typically use more electricity. When I spoke with Jeff Dean, Google’s chief scientist, he said the company doesn’t currently have plans to do this sort of analysis for images and videos, but that he wouldn’t rule it out. The reason the company started with text prompts is that those are something many people out there are using in their daily lives, he says, while image and video generation is something that not as many people are doing. But I’m seeing more AI images and videos all over my social feeds. So there’s a whole world of queries not represented here. Also, this estimate is the median, meaning it’s just the number in the middle of the range of queries Google is seeing. Longer questions and responses can push up the energy demand, and so can using a reasoning model. We don’t know anything about how much energy these more complicated queries demand or what the distribution of the range is. 2. We don’t know how many queries Gemini is seeing, so we don’t know the product’s total energy impact. One of my biggest outstanding questions about Gemini’s energy use is the total number of queries the product is seeing every day. This number isn’t included in Google’s report, and the company wouldn’t share it with me. And let me be clear: I absolutely pestered them about this, both in a press call they had about the news and in my interview with Dean. In the press call, the company pointed me to a recent earnings report , which includes only figures about monthly active users (450 million, for what it’s worth). “We’re not comfortable revealing that for various reasons,” Dean told me on our call. The total number is an abstract measure that changes over time, he says, adding that the company wants users to be thinking about the energy usage per prompt. But there are people out there all over the world interacting with this technology, not just me—and what we all add up to seems quite relevant. OpenAI does publicly share its total, sharing recently that it sees 2.5 billion queries to ChatGPT every day. So for the curious, we can use this as an example and take the company’
Next step: Stay ahead with trusted tech. See our store for scanners, detectors, and privacy-first accessories.
Original reporting: MIT Technology Review