Salesforce’s AI Energy Score helps companies choose models by energy efficiency

The rapid growth of artificial intelligence has led to mounting concerns about its energy consumption and environmental impact. Salesforce has launched a new online tool called AI Energy Score that allows users to compare and evaluate the energy efficiency of different AI models.

The big picture: The AI Energy Score tool provides standardized energy ratings and a public leaderboard to help developers and companies make environmentally conscious decisions when selecting AI models.

• The tool features a star rating system that makes it easy to identify the most energy-efficient models for common AI tasks like image creation, text generation, and note summarization
• Salesforce developed the tool in collaboration with Cohere, Carnegie Mellon University, and Hugging Face
• The system aims to bring transparency to AI energy consumption, which has been largely opaque for sustainability officers and decision-makers

Current challenges: Data center energy consumption for AI applications is projected to have significant environmental implications in the coming years.

• The Department of Energy reports that data center energy use could triple by 2028, potentially consuming up to 12% of total U.S. electricity output
• Major tech companies like Microsoft and Google have struggled to meet their climate goals due to increased energy demand from AI operations
• A December study indicates that AI growth will increase health burdens from air pollution unless data centers transition to cleaner power sources

Expert insights: Key figures involved in the project emphasize the importance of transparency and standardization in measuring AI energy consumption.

• Salesforce Senior Manager Boris Gamazaychikov highlights how sustainability officers currently struggle with the “black box” nature of external AI models and their carbon footprint impact
• Sasha Luccioni, AI and climate lead at Hugging Face, compares the tool to the Energy Star rating system for appliances, noting it provides “apples to apples” comparisons for specific AI tasks
• The standardized framework may encourage tech companies to be more transparent about their energy consumption data

Impact on industry practices: The introduction of AI Energy Score represents a significant step toward more sustainable AI development and deployment.

• The tool enables companies to align their AI usage with climate and sustainability targets
• By providing clear efficiency metrics, the system could incentivize AI developers to prioritize energy efficiency in their models
• The public leaderboard creates healthy competition for developing more energy-efficient AI solutions

Looking ahead: While AI Energy Score marks an important milestone in addressing AI’s environmental impact, its success will largely depend on industry adoption and whether it can drive meaningful changes in how companies develop and deploy AI technologies. The tool’s effectiveness in promoting transparency and efficiency could set a new standard for responsible AI development.