Generative AI, which produces complex data like language and images, is growing in popularity, but it’s raising concerns about its environmental impact. AI models require significant amounts of energy to create and use, and larger models like GPT-3 consume considerable amounts of electricity and generate carbon dioxide emissions. The energy used to manufacture computing equipment and constantly update the models also adds to their overall carbon footprint.
As chatbots and image generators become more prevalent, data centers could consume even more energy, which is why it’s important to make generative AI more efficient. Models can run on renewable energy and computation can be scheduled during times when renewable energy is more available, lowering emissions by up to 40%.
Furthermore, companies and research labs should publish the carbon footprint of their AI models, as some are already doing, and there may someday be a way for consumers to choose greener chatbots based on environmental impact.
While a single large AI model won’t ruin the environment, the collective energy use of thousands of companies using slightly different AI bots could significantly raise environmental concerns. Therefore, it’s crucial that researchers focus on making generative AI as energy-efficient as possible so that it can be used without compromising the health of the planet.