ChatGPT, the large language model developed by OpenAI, has been gaining increasing attention for its remarkable natural language processing capabilities. However, the development and training of such an AI system requires a tremendous amount of energy and high costs, resulting in significant environmental issues. In this article, we look at what we know about the environmental impact of ChatGPT and the ways we can promote responsible and sustainable practices in AI development.
Data centres are the foundation of the cloud computing needed to power AI models such as ChatGPT. Estimates suggest that ChatGPT results in 8.4 tons of carbon dioxide emission per year, more than twice the amount an individual is estimated to cause in a year. Of course, this figure is dependent on the type of power source used, with coal or natural gas-fired plants causing higher emission levels compared to renewable sources such as solar, wind, or hydroelectric power.
In addition to the carbon footprint of AI systems, a study by the University of California, Riverside, brought to light their significant water footprint. Microsoft’s training of GPT-3 alone is estimated to have consumed 700,000 litres of water, which is enough to make either 370 BMWs or 320 Teslas. Even the simple process of making conversation through ChatGPT requires the equivalent of a 500ml water bottle.
As language models continue to grow in size, it is essential to discuss ways to mitigate their negative impacts on the environment. Microsoft, for example, has acknowledged their responsibilities in climate change, and OpenAI have said that they have taken steps to reduce their carbon footprint by working closely with the team at Azure. Researchers have also developed methods and tools to assist with reporting energy and carbon usage, which can help with the accountability of the industry and promoting eco-friendly practices.
Individuals can also help by choosing to reduce the hype surrounding AI systems and by keeping track of their limitations. It is important to remember that AI does have societal implications and that we must think about the trade-offs between energy usage and performance when moving forward.
OpenAI is a research laboratory focusing on the development of artificial intelligence and its deployment in the real world. Founded in 2015 and headquartered in San Francisco, OpenAI seeks to create beneficial artificial general intelligence (AGI) while avoiding potential safety issues. OpenAI works on developing a wide range of AI-related topics, including natural language processing, reinforcement learning, and robotics. The current CEO is Sam Altman, a well-known venture capitalist, who believes in the potential of AI to positively shape the future of humanity.
Tess Posner is an American artificial intelligence researcher and the Director of AI4ALL, a non-profit organization focused on training the next generation of AI leaders. She has spent years researching AI, machine learning and natural language processing, with a focus on making AI technology more accessible and creating a more inclusive AI industry. Posner has written numerous articles on AI, as well as spoken at several industry-leading conferences. She is committed to making contributions to the field of artificial intelligence and continuing to promote sustainability within the space.
