Trustworthy and Robust AI Collaboration (TRAC) Workshop

Trustworthy and Robust AI Collaboration (TRAC) Workshop

About

The Trustworthy and Robust AI collaboration (TRAC) between MIT CSAIL and Microsoft Research is working towards fostering advances on robustness and trustworthy AI, which spans safety & reliability, intelligibility, and accountability. The collaboration seeks to address concerns about the trustworthiness of AI systems, including rising concerns with the safety, fairness, and transparency of technologies.

The 3rd TRAC workshop will be run as a virtual event. It will feature a number of short talks providing an update on the TRAC collaborations, as well as an opportunity to do deep dives in the projects below:

  • Stefanie Jegelka, ML with theoretical guarantees
  • Aleksander Madry, Towards ML you can rely on
  • Ankur Moitra, Robustness meets Algorithms
  • Daniela Rus, Efficient and Explainable ML Algorithms using Coresets
  • Tamara Broderick, Bayesian ML: uncertainty and robustness at scale
  • Vinod Vaikuntanathan, Yael Kalai, Distributed, Private and Efficient Machine Learning
  • Martin Rinard, Shuvendu Lahiri/Madan Musuvathi, State-Based Approaches for Verifying and Testing Neural Networks
  • Mohammad Alizadeh, Siddhartha Sen, Safe Online Reinforcement Learning in Networked Systems
  • Cathy Wu, Alekh Agarwal/Adith Swaminathan, Off-policy evaluation for risk-aware autonomous systems
  • John Guttag, Eric Horvitz, Exploration of robust machine learning for high-stakes predictions

Information on The Trustworthy and Robust AI collaboration (TRAC) between MIT CSAIL and Microsoft Research along with past workshops can be found here.

Microsoft’s Event Code of Conduct

Microsoft’s mission is to empower every person and every organization on the planet to achieve more. This includes virtual events Microsoft hosts and participates in, where we seek to create a respectful, friendly, and inclusive experience for all participants. As such, we do not tolerate harassing or disrespectful behavior, messages, images, or interactions by any event participant, in any form, at any aspect of the program including business and social activities, regardless of location.

We do not tolerate any behavior that is degrading to any gender, race, sexual orientation or disability, or any behavior that would violate Microsoft’s Anti-Harassment and Anti-Discrimination Policy, Equal Employment Opportunity Policy, or Standards of Business Conduct. In short, the entire experience must meet our culture standards. We encourage everyone to assist in creating a welcoming and safe environment. Please report any concerns, harassing behavior, or suspicious or disruptive activity. Microsoft reserves the right to ask attendees to leave at any time at its sole discretion.

Agenda

Each day’s agenda starts at 9:30am PDT sharp.

Monday, June 22, 2020

Time (PDT) Session Speaker / Talk Title
9:30-9:35 Welcome and Introduction Aleksander Mądry, Massachusetts Institute of Technology and Evelyne Viegas, Microsoft

Eric Horvitz, Microsoft and Daniela Rus, Massachusetts Institute of Technology

9:35-10:10 Opening Talk Constantinos Daskalakis, Massachusetts Institute of Technology
Robust Learning from Censored Data

Watch this session on-demand

10:10-10:15 Break
10:15-11:05 Lightning Talks Tamara Broderick, Massachusetts Institute of Technology
Approximate Cross-Validation for Complex Models

Ankur Moitra, Massachusetts Institute of Technology
Classification under Misspecification, and Implications for Fairness

Maggie Makar, Massachusetts Institute of Technology
Learnability of Contagious Infections Under Incomplete Testing

Kai Jia, Massachusetts Institute of Technology
Exploiting Verified Neural Networks via Floating Point Numerical Error

Lucas Liebenwein, Massachusetts Institute of Technology
Provable Filter Pruning for Efficient Neural Networks

Watch this session on-demand and Tamara Broderick’s Lightning Talk

11:05-11:15 Break
11:15-12:15 Breakout Sessions Select One:

  • Real-world ML Safety and Reliability
  • Coping with Biases in Data, with a Focus on ML for Healthcare (including the COVID-19-related efforts)
12:15-12:30 Break
12:30-1:00 Report Out Watch this session on-demand

Tuesday, June 23, 2020

Time (PDT) Session Speaker / Talk Title
9:30-9:35 Welcome and Introduction Aleksander Mądry, Massachusetts Institute of Technology and Evelyne Viegas, Microsoft

Eric Horvitz, Microsoft and Daniela Rus, Massachusetts Institute of Technology

9:35-10:10 Opening Talk Hal Daumé III, Microsoft
Language (technology) is Power

Watch this session on-demand

10:10-10:15 Break
10:15-11:05 Lightning Talks Yael Kalai, Microsoft
Learning with Arbitrary Adversarial Test Examples

Stefanie Jegelka, Massachusetts Institute of Technology
Unsupervised Risk Estimation with Domain-Invariant Predictors

Hadi Salman, Microsoft
Do Adversarially Robust ImageNet Models Transfer Better?

Pouya Hamadanian, Massachusetts Institute of Technology
Towards Safe Online Reinforcement Learning in Computer Systems

Cathy Wu, Massachusetts Institute of Technology
Policy transfer across networks: towards understanding AI impacts in urban-scale systems

Watch this session on-demand

11:05-11:15 Break
11:15-12:15 Breakout Sessions Select One:

  • Human-ML Interface: Interpretability/Explainability
  • Adversarially Robust ML
12:15-12:30 Break
12:30-12:50 Report Out
12:50-1:00 Closing Remarks and Next Steps Eric Horvitz, Microsoft and Daniela Rus, Massachusetts Institute of Technology

Aleksander Mądry, Massachusetts Institute of Technology and Evelyne Viegas, Microsoft

Watch this session on-demand

Videos