A panel discussion around what is "Out of Distribution" (OOD) problem and why is the OOD problem important?

Panelists: Yoshua Bengio, Leyla Isik, Max Welling

Moderators: Joshua Vogelstein, Weiwei Yang

Join us for live Q&A with our distinguished panelists where attendees can pose questions to the panelists and engage in dialogue.

Please feel free to post questions prior to the session in rocket.chat and we will do our best to answer them during the session.

Panelists: Yoshua Bengio, Leyla Isik, Konrad Kording, Bernhard Scholkopf, Amit Sharma

Moderators: Joshua Vogelstein, Weiwei Yang

Join us for an extension of our Live Q&A Session 1 and Learning Salon, where we will continue to discuss and answer attendee questions!

Panelists: Anima Anandkumar, Chris White, Terry Sejnowski

Moderators: Joshua Vogelstein, Weiwei Yang

Questions:

• What does the "Out of Distribution" (OOD) problem mean to you or your field?
• Why is the OOD problem important?
• What insights did you gain from previous work that might help with solving the OOD problem and successfully applying it to tackle real world challenges?
• What do you think the ideal role of AI is in our life?
• What set of achievements or breakthroughs need to happen to bridge the gap between ML's current role and its ideal role?

Join us for live Q&A with our distinguished panelists where attendees can pose questions to the panelists and engage in dialogue.

Please feel free to post questions prior to the session in rocket.chat and we will do our best to answer them during the session.

Panelists: Susan Athey, Yoshua Bengio, Sujeeth Bharadwaj, Jane Wang

Moderators: Joshua Vogelstein, Weiwei Yang

Panelists: Doug Burger, Alysson Muotri, Ralph Etienne-Cummings, Florian Engert

Moderator: Soledad Villar, Teresa Huang

Topics for discussion:

• Learning with constraints in artificial intelligence (AI) and natural intelligence (NI): definition and examples
• Benefits from constrained-based learning: data, energy, time
• Experimental design of constrained-based learning in AI and NI

Panelists: Susan Athey, Konrad Kording, Amit Sharma

Moderator: Emre Kiciman

Opening Question

• In everyday life, understanding the causal mechanics of the world around us seems to be -- at least at first glance --- incredibly intuitive to most people. And yet, like many things that people find to be "common sense", understanding and capturing causality is also often surprisingly difficult to operationalize. My first question for the panel --- could you say a few words about causality and its importance in your areas of study, what excites you, and what are the open challenges you see?
• Konrad, what are some of the things that people do well when reasoning about causality. What do we know about how we do that?
• Susan, your work often involves policy setting and decision making where people must understand the mechanisms underlying a market, or other system. Where do you see people making mistakes? How do the formal analyses you work with complement people's cognitive abilities?
• Amit, in your work integrating causal concepts into computing applications, what are the opportunities you see for new (biologically-inspired) models to have an impact on our algorithms and computing methods?
• Each of you takes a broad cross-disciplinary approach in your research---re-reading through your backgrounds before the panel, I was reminded of the breadth of work here, spanning at least large segments of computer science, neuroscience, and social science. With your perspectives at interdisciplinary boundaries, what are some of the key touchpoints that you see as opportunities at the intersection of natural and artificial intelligence?

Panelists: Judith Fan, Kevin Ellis, Brenden Lake, Josh Tenenbaum

Moderator: Zenna Tavares

Question:

• How has understanding of compositionality today has changed relative to more classical theories? Given what we know today in the field of cognitive science and AI, how should we update our understanding of what compositionality means?
• Is compositionality something of the world that we're trying to extract or is it in our minds that we're trying to impose onto the world?
• Who or what is compositionality for?
• How are abstraction and compositionality related?
• What should we be thinking about with respect to compositionality as we go into the future?

Panelists: Celeste Kidd, Chelsea Finn, Timothy Verstynen

Moderator: Johnathan Flowers

Questions:

• How do you define curiosity as it relates to Machine Learning and AI?
• What, if any, is the value of developing a curious machine or AI?
• How might we approach developing a curious AI?
• What risks are associated with Curiosity in these projects?

Panelists: Jane Wang, Eric Traut, Gabriel Silva, Sujeeth Bharadwaj

Moderator: Weiwei Yang

Questions:

• How would you define continual learning?
• What are the most important mechanisms that allow a system (biological, artificial or hybrid) to learn continuously?
• What is the future of AI?