Computer Vision News - April 2022

25 Claire Vernade debate lately in machine learning, very broadly, about the quality of evaluation, the validity of theoretical results, even just the validity of empirical results. I find that people don’t dare ask too many theoretical questions because they sound like fundamental theory questions in machine learning, that people should not ask about. In fact, this is a very difficult learning theory: this is a whole part of machine learning. There are people working on it full time and it’s difficult. We notice at DeepMind that opening a Slack channel called “ Ask Foundations ” has had so much success! People come and ask very basic questions. I want to say that people should feel comfortable asking questions even when they seem basic, even fundamental questions that sound like graduate school questions. Very often they aren’t. Let’s exchange! Let’s ask questions!  Over the past two years, I’ve become very keen to learn about the economy. It’s in the news, and everyone talks about it. The currencies and everything are so technical. There is so much to learn about the economy.  What would you change in economics?  The question is very broad. It’s very political. People should never forget that everything is political. This is one thing that I’ve seen looking at economic theories. Sometimes they tendto forget that some of the decisions or models affect peoples’ lives. These are decisions that, in democracies, we have to make together.  Do you have a final message to share with our community? Wow! There has been a lot of I enjoy a lot looking at real-world problems and asking the question: How can I write it in math? How can I translate it into a model that I can solve?