Project ISF: Uncertainty Quantification in Life Cycle Assessment for Sustainable Computing Program DUIRI - Discovery Undergraduate Interdisciplinary Research Internship Term Fall 2024 Status Accepted Research Area Sustainable Computing, Artificial Intelligence Description The widespread adoption of Large Language Models (LLMs) in industry such as ChatGPT has led to an exponential increase in computing demands and energy consumption, posing sustainability challenges. For instance, serving one prompt in ChatGPT generates more than 4 grams of carbon emissions — over 80 times the carbon emission of a web search query. The development of LLMs not only demands significant energy, but also requires substantial computing hardware resources. To mitigate the environmental impact of LLMs, a thorough understanding of its carbon emissions across various hardware platforms is essential. This project aims to characterize the environmental impact and their uncertainty levels across various hardware platforms, enhancing our understanding of life cycle assessment for sustainable computing. Supervisor Yi Ding Mentor Inez Hua Type of work required Students are expected to dedicate 10-15 hours per week to this project. They will participate in weekly meetings with faculty members and graduate student mentors to report their progress. Key responsibilities include:
1. Understanding life cycle assessment and applying it to assess the environmental impact of modern AI systems, such as LLM training and serving systems.
2. Utilizing existing tools like PAIA to quantify the uncertainty levels of embodied carbon footprints.
3. Characterizing and understanding the trade-offs between carbon footprints, performance, and costs.
4. Building datasets and documenting results for publication.

Websites Qualifications Engineering major at Purdue-West Lafayette, junior or senior level, with interest in artificial intelligence and sustainable computing. Programming skills are preferred. Credit Hours 0 Weekly Hours 10 (estimated)