Microsoft has made an ambitious commitment to remove its carbon footprint in response to the overwhelming urgency of addressing climate change. Meanwhile, recent advances in machine learning (ML) models, such as transformer-based NLP, have produced substantial gains in accuracy at the cost of exceptionally large compute resources and, correspondingly, carbon emissions from energy consumption. Understanding and mitigating the climate impact of ML has emerged at the frontier of ML research, spanning multiple areas including hardware design, computational efficiency, and incentives for carbon efficiency.
The goal of this session is to identify priority areas to drive research agendas that are best-suited to efforts in academia, in industry, or in collaboration. We aim to inspire research advances and action, within both academia and industry, to improve the sustainability of machine learning hardware, software and frameworks.
Session Lead: Philip Rosenfield, Microsoft
Speaker: Nicolo Fusi, Microsoft
Talk Title: Opening Remarks
Speaker: Emma Strubell, Carnegie Mellon University
Talk Title: Learning to Live with BERT
Speaker: Vivienne Sze, Massachusetts Institute of Technology
Talk Title: Reducing the Carbon Emissions of ML Computing – Challenges and Opportunities
Speaker: Diana Marculescu, University of Texas at Austin
Talk Title: When Climate Meets Machine Learning: The Case for Hardware-ML Model Co-design
Q&A panel with all 4 speakers