The Weather Network reached out to Microsoft, Google and OpenAI for comment. Microsoft declined to say how much energy will be used to train the Bing implementation of GPT-4, which could replace Bing search. He declined to comment specifically on how the introduction of these AI models would affect the company’s carbon footprint. A Google spokesperson declined to answer a similar question. OpenAI did not respond to comments.
“As part of our commitment to creating a more sustainable future, Microsoft is working on ways to make large-scale systems more efficient in both training and applications, while reducing the energy use and carbon impact of AI. We invest in research that measures,” a Microsoft spokesperson said. Said. “We also continue to invest in renewable energy purchases and other initiatives to reach our sustainability goals of carbon negative, water positive and zero waste by 2030.”
table from Artificial Intelligence Index Report 2023.
“Energy usage is determined by many factors, including the size of the data the system is trained on and the energy efficiency of the data center the system is trained on,” Masley said. “All else being equal, if the observed trend toward ever-larger deployments of AI models continues, the amount of energy required to train them is likely to continue to increase. I have.”
Model parameters went from 100 million in 2018 to 500 billion in 2023.
These AI models don’t stop at training. It takes more energy to upload models to the internet and keep them online for people to access. This allows you, for example, to ask ChatGPT to perform tasks or answer questions. Each time the model receives a new prompt, it requires an additional source of energy to respond.
