Policies for sustainable development can entail complex decisions about balancing environmental, economic, and societal needs, and making these decisions in an informed way presents significant computational challenges.  Modern AI techniques combined with new data streams have the potential to yield accurate, inexpensive, and highly scalable models to inform research and policy. At the Deep Learning Summit in San Francisco, Stefano Ermon, Assistant Professor at Stanford University, will explore this further, presenting an overview of his group's research on applying computer science techniques in sustainability domains, including poverty and food security. Stefano is affiliated with the Artificial Intelligence Laboratory and the Woods Institute for the Environment, where he centres his research on techniques for scalable and accurate inference in graphical models, statistical modeling of data, large-scale combinatorial optimization, and robust decision making under uncertainty. His research was recently featured in National Geographic, after he and his team successfully combined satellite data and machine learning to map poverty. I asked him a few questions ahead of the summit to learn more.What started your work in machine learning? Why were you motivated to work in this field? I studied control theory as an undergrad - I've always been fascinated by the idea of building systems that can make smart decisions in real world environments. Machine learning was the natural next step as I was thinking about increasingly complex scenarios and reasoning patterns. I was fascinated by the rapid progress of the discipline, and I wanted to make my contribution.  Computational sustainability is a rapidly growing field. What are current factors for this? There is a huge demand for more computational thinking in sustainability disciplines. Our sensing capabilities (from satellites to smartphones and citizen science projects) are booming and we are producing enormous amounts of data. We desperately need better ways to extract useful insights from this data and help us make better decisions about some key challenges, from sustainable energy to biodiversity conservation, that impact all of us.   In which areas do you see the biggest value for applying machine learning? Machine learning is a great tool to build computer systems that interact with physical environments (and humans). It's a powerful framework that deals quite naturally and robustly with uncertainty, misspecifications and the "fuzziness" of the real world. Natural application domains that come to mind are healthcare, education, and sustainability sciences. I also see a lot of unexplored potential in traditional scientific disciplines, like physics and chemistry.  What developments can we expect to see in machine learning in the next 5 years? It's hard to make predictions in a field that is moving so fast, but I think we're going to see more integration between the recent amazing results in key perceptual tasks (vision, speech, etc.) with more traditional decision-making and reasoning frameworks from AI.  Outside of your own field, what area of deep learning advancements excites you most? I'm pretty excited about some of the results I've seen in applying deep learning in biology and genomics. It's an area where collecting data is becoming cheaper and cheaper - combined with more powerful ML models, the potential is huge.Stefano Ermon will be speaking at the Deep Learning Summit in San Francisco on 26-27 January. Other speakers include Ilya Sutskever, Research Director at OpenAI; Lise Getoor, Professor of Computer Science at UC Santa Cruz; Naveen Rao, CEO & Co-Founder of Nervana Systems; Danny Lange, Head of Machine Learning at Uber; and Brendan Frey, Co-founder & CEO of Deep Genomics.Early Bird passes for the Deep Learning Summit end on 7 October. Book your pass now on the website here. You can also learn more about the medical applications of AI at the Deep Learning in Healthcare Summit, taking place in Boston and London. See the full events list here.