Clare Kanazawa is a student at Holy Names Academy in the United States. She is a member of the National Honor Society and has been recognised with the Sr. Ann Book Award and a varsity robotics letter. Clare is actively involved in robotics as a programming team member, serves as a Youth Ocean Advocate at the Seattle Aquarium through its Leadership Pod, and participates in the Art of Problem Solving programme for calculus. She is also a section leader in her school orchestra, performing in piano and percussion.
At the CCIR Academy, Clare’s research paper, Source Attribution vs. Non-Source Attribution in Large Language Models: An Analysis of Which Technique Is More Environmentally Friendly, has been accepted at the Sustainable AI Conference. Her research compares the environmental costs of source-attributed and non-source-attributed large language models. By fine-tuning LLaMA models and measuring energy use and carbon emissions, the study finds that source attribution leads to higher computational energy consumption and emissions, suggesting that reduced attribution may help lower the environmental footprint of large language models.
Hosted by the Bonn Sustainable AI Lab and the Institute for Science and Ethics at the University of Bonn, the Sustainable AI Conference brings together researchers, policymakers, and industry leaders to examine the environmental and societal impacts of artificial intelligence. The conference provides a platform for discussing approaches to developing and using AI in more sustainable and responsible ways.
Congratulations to Clare on this achievement!
