Do AI language models like ChatGPT harm the environment due to high energy consumption?

Generative AI, a powerful technology used in various fields, has been found to have a significant carbon footprint. A recent study by researchers at Hugging Face and Carnegie Mellon University revealed that generating a single image using AI consumes as much energy as fully charging a smartphone. The study analyzed the impact of 10 prompts on 88 different AI models and found that creating images was the most carbon and energy-intensive activity.

Using a powerful AI model like Stable Diffusion XL to create 1,000 images generates approximately the same amount of carbon dioxide as driving an average petrol-powered car for 4.1 miles, or about 1.1 kilograms of CO₂. However, text generation models have a significantly lower carbon footprint, with the lowest-carbon model producing only as much CO₂ as a 0.0006 mile journey in a similar vehicle, or around 0.002 kilograms of CO₂.

It is worth noting that the high CO₂ emissions of AI imagery do not arise solely from the energy used for training large models on supercomputers. In fact, the majority of emissions result from the actual use of these models. For instance, the energy used to train large language models like ChatGPT is exceeded after just a few weeks of use, considering that ChatGPT has around 10 million daily users. Studies also suggest that running large generative models is significantly more energy-intensive than using more specific models designed for specific tasks.

The researchers involved in the study hope that these findings will drive more conscious consumption of generative AI and the selection of energy-efficient models whenever possible. They also aim to raise awareness of this issue and urge companies to take responsibility for their energy footprint.