- A new study reveals that every query to large language models like ChatGPT consumes energy and generates carbon emissions.
- Complex reasoning questions, such as those in abstract algebra or philosophy, lead to significantly higher carbon emissions, up to six times more than simpler queries.
- Models designed for explicit reasoning processes produce substantially more carbon dioxide, with some generating up to 50 times more emissions than concise response models.
- The study highlights an "accuracy-sustainability trade-off," where highly accurate AI models often result in greater energy consumption and carbon footprint.
- Researchers recommend that users reduce emissions by prompting AI for concise answers and reserving high-capacity models for tasks that genuinely require their advanced capabilities.
IN FULL
