- Google released a study that outlined how much energy and water it uses
- Experts say the numbers lack important context
- Google insists that these numbers represent the average user experience
A new study from Google claims that the Gemini AI model uses only a lot of minimal water and energy for each prompt -with the median consumption, which sits at about 5 drops (0.26 ml) -equivalent electricity used for 9 seconds of TV monitoring (about 0.24 Watt hours), resulting in about 0.003 grams of CO2 emissions.
However, experts have been quick to contest the requirements The Verge To require Google omitted key data points in its study, drastically under -reporting the environmental impact of the model.
While models and data centers have become more effective, it seems that there is more to the story than Google seems.
The tip of the iceberg
One of the authors of a paper cited in the study, Shaolei Ren, Associate Professor of Electrical Technology at the University of California told the publication; “They just hide the critical information. This really spreads the wrong message to the world.”
AI models such as Gemini are supported by data centers – huge stocks full of servers that consume intense amounts of water and energy, and strain local resources.
Governments across the globe have sanctioned the building of these data centers despite the destruction they could bring to the local landscape – and consumers are probably those who pay for the extra energy used.
One of the biggest concerns with Google’s study is that it omits indirect water consumption in the estimates that form most of the use related to AI. While the figures are technically correct, the lack of context of the extreme energy consumption paints a misleading image.
The study only looks at the water used by data centers to cool their servers, but omitted is the electricity that these data centers require, which in turn leads to new gas and nuclear systems – which also cools their systems with water or uses steam to turn turbines.
However, water is not the only metric Google error -like represented, with the paper sketching only a ‘market -based’ carbon emission measure that offsets the figure that uses Google’s promises to use renewable energy to support power networks. Savannah Goodman, head of Advanced Energy Labs told Techradar Pro,
“We hope to share environmental measurements that are representative of a typical user’s behavior and reasonably comparable over time. However, with the rapidly evolving landscape of AI model architectures and AI assistant user behavior, there are outliers either from small subgroups of prompt served by low utilization models or with high token counts.”
“To share measurements that represent a typical user’s experience and are robust to this rapidly evolving field, we chose to measure measurements for the median promotion – which is robust to extreme values and provides a more accurate reflection of a typical prompt’s energy impact.”
