The study revealed that generating images was the most energy- and carbon-intensive task, with 1,000 images equating to the CO2 emissions of driving 4.1 miles in a gasoline-powered car. The researchers also found that using large generative models was far more energy-intensive than using smaller, task-specific models. The study's findings could encourage more selective use of AI and the development of less carbon-intensive models. The researchers also found that the day-to-day emissions from using AI far exceeded the emissions from training large models.
Key takeaways:
- Generating an image using a powerful AI model takes as much energy as fully charging a smartphone, while creating text 1,000 times only uses as much energy as 16% of a full smartphone charge, according to a study by Hugging Face and Carnegie Mellon University.
- Most of the carbon footprint of AI models comes from their actual use, not just their training, and using large generative models is far more energy-intensive than using smaller, task-specific models.
- Generating 1,000 images with a powerful AI model is responsible for roughly as much carbon dioxide as driving the equivalent of 4.1 miles in an average gasoline-powered car.
- The day-to-day emissions associated with using AI far exceeded the emissions from training large models, with popular models such as ChatGPT potentially having their usage emissions exceed their training emissions in just a couple of weeks.