Interview with Luxuan Sun

Sun’s research focuses on ICT-driven consumer industries and how they shift emissions across the full system, using life cycle assessment. This focus and emphasis on life cycle assessment directly translates to some of the proposed solutions surrounding AI’s environmental impact, as it shows the full picture of a technology. 

Sun received her bachelor’s degree from Chongqing University and her master’s from Shanghai Jiao Tong University, so I asked her about a cross-country comparison between China and the United States. She believes that energy structure is often the most important factor in determining the environmental impact of AI in both countries. She recommends embedding grid mix assumptions directly into a system model.

She notes that a typical misconception people have is surrounding the differences between carbon and GHG emissions. It is important to be explicit about whether it is CO₂ only or broader GHGs in CO₂e. 

The typical research challenges she faces are in data collection and validation because of sensitivity to workload definition, utilization, and cooling estimations. Additionally, it is useful to separate data that is operator-reported versus estimated–being explicit matters here.