Bill Tomlinson, Rebecca Black, Andrew Torrance, and I wrote a paper that was accepted to Springer Nature Journal, Scientific Reports, in which we compare the carbon emissions between AI systems and humans.
What did we find?
- AI systems emit significantly less CO2e (carbon dioxide equivalent) per page of text generated and per image created compared to human writers and illustrators. Specifically, AI systems emit between 130 and 1500 times less CO2e per page of text and between 310 and 2900 times less CO2e per image.
- We acknowledges that while AI has a lower carbon footprint for these tasks, it does not account for social impacts such as professional displacement, legality, and rebound effects, nor is AI a substitute for all human tasks.
- Despite these considerations, the current use of AI in writing and illustration tasks offers the potential for significant reductions in carbon emissions compared to human performance.
Rapid advancements in AI and its applications across various domains, raise concerns about environmental impact, particularly in terms of energy consumption and greenhouse gas emissions. This study aimed to compare the environmental impact of AI and humans in writing and illustration tasks, contributing to discussions on sustainable consumption and production patterns
We conducted numerical analyses to assess the environmental impacts, focusing on the energy consumption and carbon emissions of AI systems (ChatGPT, BLOOM, DALL-E2, Midjourney) and human activities involved in writing and illustrating tasks.
We discuss the benefits and drawbacks of AI, including potential job displacement and legal issues, while highlighting the lower environmental footprint of AI for certain tasks. The paper argues for collaboration between AI and human labor, leveraging their respective strengths for efficient and environmentally sustainable outcomes.
The findings are based on current technology and societal conditions, with future changes likely to affect the environmental impact of both AI and human activity.
Related papers by other (also see references in the paper):
Potential reduction in healthcare carbon footprint by autonomous artificial intelligence
(permanent , open-access, local copy)
C.V.: JR-18