Stefano Ermon Interview on This Week of Machine Learning & AI

Domain Knowledge in Machine Learning Models for Sustainability with Stefano Ermon

800 800 The TWIML AI Podcast (formerly This Week in Machine Learning & Artificial Intelligence)

My guest this week is Stefano Ermon, Assistant Professor of Computer Science at Stanford University, and Fellow at Stanford’s Woods Institute for the Environment. Stefano and I met at the Re-Work Deep Learning Summit earlier this year, where he gave a presentation on Machine Learning for Sustainability.

Stefano and I spoke about a wide range of topics, including the relationship between fundamental and applied machine learning research, incorporating domain knowledge in machine learning models, dimensionality reduction, semi-supervised learning, proxy & transfer learning, and his interest in applying ML & AI to addressing sustainability issues such as poverty, food security and the environment.

New Stickers, New Contest

This Week in Machine Learning & AI StickerInspired by an exchange with listener @bethann_nyc on Twitter, we want to try a little contest for you all. We’ve got some fresh new stickers on hand, and we’d like to get them into your hands! (I’ve got to say, they really turned out great!) So here’s what you need to do to get one: Let us know your favorite quote from this week’s podcast via Facebook, Twitter, Youtube or SoundCloud comment, or a comment on the show notes page by midnight Sunday (3/19) Pacific Time and we’ll send you one!

Future of Data Summit

About Stefano Ermon

Mentioned in the Interview

  • Mason Grimshaw

    “Transfer Learning is trying to find a task that correlates well with what you’re trying to figure out, and you have a lot of data for that.”

    I had never heard about this before, and it makes a lot of sense! I wonder where else people use this?

    I love the show!! Keep it up.

  • Chase

    As an environmental engineer, I was happy to hear this episode. I use ML techniques for some of my work and the lack of their ability to integrate physics can lead to issues. So it’s no surprise that my favorite quote was:

    “Ideally I would like to distill physics and knowledge from the raw data. I want my machine learning system to come up with new hypotheses and discover gravity… I think that is what is really exciting.”

  • Bill Glennon

    I just came across this podcast. I really enjoyed this one and looking forward to listening to some others. Keep up the great work!

    “…Beauty of computer science is the idea of abstraction of … develop general models, general algorithms are inspired by one specific problem then a few years down the road a new application comes from this for a different problem being solved…”

  • Daniel Aragon

    Listening to Dr. Ermon’s talk, I was intrigued by the idea of using NASA’s satellite data to predict poverty. In referencing his paper in NATURE, Dr. Ermon stated the ability for the process to “outperform previously existing techniques by a large margin.” Amazing that physical attributes collected via space can produce insight into our economic well being. Equally, this affirms that for many of us on this planet, our wealth is directly tied to earth’s resources and how we’ve co-opted them to serve our livelihood. Great show! How about that sticker?

    • sam

      I continue to refer back to that show. In fact, I mention it in the pod that goes up on Friday. Re: the sticker, of course! Thanks for commenting!

  • Dmitry

    Thank you Dr. Ermon for the very interesting and powerful work. The ability to take huge feature sets of multi-spectral data and combine it with other information is very powerful. This is all done to make the world a better place, too! A lot of great work. Thank you.

Leave a Reply

Your email address will not be published.