Source URL: https://tech.slashdot.org/story/25/08/22/2118234/google-says-it-dropped-the-energy-cost-of-ai-queries-by-33x-in-one-year?utm_source=rss1.0mainlinkanon&utm_medium=feed
Source: Slashdot
Title: Google Says It Dropped the Energy Cost of AI Queries By 33x In One Year
Feedly Summary:
AI Summary and Description: Yes
Summary: Google’s recent analysis reveals a significant reduction in the energy consumption and carbon emissions associated with its AI text queries, achieving a 33x reduction over the past year. This improvement is attributed to software optimizations and advancements in renewable energy usage, highlighting the intersection of AI advancements with environmental considerations.
Detailed Description:
* Google has released a new analysis regarding the environmental impact of its AI technologies, particularly focusing on the energy use associated with AI text queries.
* Key findings include:
– A reduction in energy consumption from text queries by a factor of 33 over the past year.
– Each prompt now uses about 0.24 watt-hours of energy, comparable to the energy consumed during nine seconds of television viewing.
– The carbon dioxide equivalent emitted per query is about 0.03 grams, showcasing a relatively low individual environmental impact despite the high volume of queries processed.
* Contextual considerations:
– Although the energy usage is minimal for individual queries, the cumulative environmental impact is likely substantial due to the high number of requests.
– The growth of renewable energy sources, particularly solar power, has facilitated a decrease in carbon emissions per unit of energy consumed (a 1.4x reduction over the past year).
* Key optimizations made by Google contributing to energy efficiency include:
– Implementation of the Mixture-of-Experts approach, which activates only the necessary parts of the AI model to significantly reduce computational demands (by a factor of 10 to 100).
– Development of compact model versions that lessen the computational load further.
– Enhanced data center management strategies ensuring optimal utilization of active hardware and maintaining low-power states for idle equipment.
* Hardware and Software Optimization:
– Google has designed custom AI accelerators and optimized the software for these accelerators, aligning hardware capabilities with software requirements. This integration is critical, as it accounts for over half of the total energy used per query.
– The company’s extensive experience in operating efficient data centers has been leveraged to enhance AI performance.
In summary, Google’s initiatives not only signify advancements in AI efficiency but also reflect growing awareness and responsibility towards environmental sustainability within the tech industry, indicating a broader trend of integrating environmental considerations within AI development and operational frameworks.