Generating an AI image consumes as much energy as charging a smartphone

Occurred: November 2023

Creating an image with generative AI uses as much energy as charging your smartphone, according to a new research study.

Using a tool called Code Carbon, the researchers from Carnegie Mellon University and Hugging Face compared consumption and emission data across 16 of the most popular models found on HuggingFace Hub, and found that general-purpose AI models mostly consume much more energy than single purpose models. The researchers also found that text-based AI tasks are more energy-efficient than jobs involving images.

The findings highlight the high energy consumption demands of generative AI systems and particularly text-to-image models, and increassed pressure on generative AI developers to be more transparent about the environmental costs of their technologies.

Operator: Sasha Luccioni; Yacine Jernite; Emma Stubell
Developer: Midjourney; OpenAI
Country: Global
Sector: Multiple
Purpose: Generate image
Technology: Text-to-image; Generative adversarial network (GAN); Neural network; Deep learning; Machine learning
Issue: Environment
Transparency: Governance

Research, advocacy 🧮