The computational needs of the powerful machine learning (ML) models have increased several orders of magnitude in the past decade with the amount of compute doubling every six months. The estimated carbon footprint of a state-of-the-art large language model is equivalent to the average yearly carbon footprint of several households. This makes the current rate of increase in model parameters and datasets unsustainable.

To achieve more sustainable progress in ML, it is essential to invest in more resource-/energy-/cost-efficient solutions. This session will explore how to make computational and carbon footprint of ML more transparent and improve ML resource efficiency.

We plan to take a holistic view, which includes data preparation and loading, continual retraining of models in dynamic data environments, compiling ML on specialized hardware accelerators, and serving models for real-time applications with low-latency requirements and constrained resource environments.

Program

Brief introduction to the session (10mins)

Talks (~7min talk, ~3 min Q&A) from people approaching the challenge of sustainability of ML from different angles (50mins)

Speakers

• Evin Duran, Sustainability Ambassador, Oracle, Denmark
• Subham Sahoo, Assistant Professor, Aalborg University
• Ties Robroek, Postdoc, IT University of Copenhagen
• Sebastian Bugge Loeschcke, PhD student, University of Copenhagen & Pioneer Center for AI
• Emil Jørgensen Njor, PhD Student, Technical University of Denmark

Open discussion with the audience (30mins)

Organizers

• Pınar Tözün, Associate Professor, IT University in Copenhagen
• Raghavendra Selvan, Assistant Professor, University of Copenhagen

Level

Intermediate: For attendees who have basic understanding or some experience with the subject but are not yet advanced.