1:30 - 4:30pm Thursday, December 10, 2009

Regency D

This is part of the Mini Symposia which begins at 13:30 on Thursday December 10, 2009

The world has a sustainability problem. Humans currently consume an average of 16TW of power (and rising), more than 86% of which comes from (unsustainable) fossil fuels. There is a range of estimates as to when this supply will run out, but this is a scenario that may well happen within our lifetimes. Even more pressing is the effect that such fuels have on our climate: given no attempts to reduce the world's fossil fuel usage, even the most conservative climate models predict that the world temperature will increase by over five degrees (Fahrenheit) in the next 90 years, an increase that could cause ecological disasters on a global scale. Building a sustainable infrastructure for energy and ecosystems is shaping up to be one of the grand scientific and political challenges of the 21st century. Furthermore, there is a growing consensus that many aspects of sustainability are fundamentally information systems problems, tasks where machine learning can play a significant role. This mini-symposium will bring together leading researchers with both machine learning backgrounds and energy/sustainability backgrounds to address the question: how can machine learning help address the world's sustainability problem? The mini-symposium will also seek to answer: What is the current state of work directed at sustainability, energy, and ecology in the machine learning, operations research, and optimization communities? What are the primary scientific and technical challenges in information processing for sustainability? And finally, what are (and what aren't) areas where machine learning can make a genuine impact on the science of sustainability? Because this is an emerging field of research, the talks at this symposium will aimed at the general NIPS audience. There is a growing number of researchers working in sustainability, but even more broadly, we think that such problems have the potential to advance basic machine learning in a manner similar to other important applications, such as computer vision, natural language processing, and computational biology. Sustainability problems offer an equally rich set of domains, and solutions to these problems will have a genuine impact on the world.

http://www.cs.stanford.edu/group/nips09-mlsust/