Google is taking a significant step towards harnessing the power of artificial intelligence (AI) while reducing its environmental impact. The company has developed a comprehensive methodology for measuring the energy, water, and carbon emissions of its AI models. This move comes as AI becomes increasingly important in various fields such as medicine, energy, and transportation.
To better understand AI’s environmental footprint, Google is investing heavily in research and development. The goal is to create more efficient AI systems that consume less energy while maintaining their performance. Recent progress shows promise, with median energy consumption per Gemini Apps text prompt decreasing by factors of 33x and 44x over a 12-month period.
Notably, this work builds upon Google’s long-standing commitment to data center efficiency. In 2024, the company reduced its data center energy emissions by 12% despite an increase in electricity consumption. This study aims to contribute to ongoing efforts to develop efficient AI that benefits everyone, particularly in addressing critical issues like energy sustainability and scientific discovery.
To achieve this goal, Google is adopting a full stack approach to energy innovation, including new infrastructure investments, grid engineering, and clean energy solutions. The company’s work on model efficiency is also crucial in minimizing the environmental impact of AI. With transparency and collaboration being key drivers of progress, Google invites readers to learn more about its research and technical report.
Source: https://blog.google/outreach-initiatives/sustainability/google-ai-energy-efficiency