Artificial intelligence tools, which have become increasingly common in recent years, have brought to the fore the issue of carbon footprint, which many of us perhaps ignore. Although it is useful for us in many areas, the energy consumed by artificial intelligence is considered a harbinger of disaster for a certain segment of people.
However, Hugging Face, an initiative project on artificial intelligence, together with Carnegie Mellon University, conducted a research on the energy consumed by artificial intelligence vehicles. Especially while DALL-E created incredibly realistic images, the energy it spent attracted the attention of many people.
According to research, the energy consumed by artificial intelligence tools when creating images is equivalent to charging a phone.
DALL-E, perhaps the most popular of the rendering tools, uses only the energy used to charge a phone up to 16% when rendering. Accordingly, the energy consumed by artificial intelligence does not increase its carbon footprint that much.
Moreover, text-based artificial intelligence tools such as ChatGPT consume less energy than the energy we talk about when summarizing an article or chatting about any topic. This shows how small the carbon footprint actually is.
A total of 13 different commands were examined in the research.
A total of 13 different commands were discussed, including not only creating images and text, but also text summarization and classification. Then, they examined the amount of carbon dioxide produced per 1000 grams and worked on 88 different models with 30 different data sets to make the study more fair.
The researchers ran a total of 1,000 commands to obtain the amount of carbon during the measurement. Accordingly, the tools that need the most energy are listed as text creation, summarization and image creation. Image rendering ranked first in terms of the emissions it produced, while text rendering ranked last.
ChatGPT, which has more than 10 million daily users, has caused researchers to conduct research on carbon footprint. Researchers invite developers to be transparent in order to make the impact of artificial intelligence on the environment more understandable. Although they emphasize that charging a phone up to 16% consumes a small amount of energy, carbon emissions can easily increase.
51027
