Offline data are both valuable and practical resources for teaching robots complex behaviors. Ideally, learning agents should not be constrained by the scarcity of available demonstrations, but rather generalize beyond the training distribution. However, the complexity of real-world scenarios typically requires huge amounts of data to prevent neural network policies from picking up on spurious correlations and learning non-causal relationships. We propose CAIAC, a data augmentation method that can create feasible synthetic transitions from a fixed dataset without having access to online environment interactions. By utilizing principled methods for quantifying causal influence, we are able to perform counterfactual reasoning by swapping $\it{action}$-unaffected parts of the state-space between independent trajectories in the dataset. We empirically show that this leads to a substantial increase in robustness of offline learning algorithms against distributional shift.

A coalition of AI groups is forming to produce a comprehensive data source on the coronavirus pandemic for policymakers and health care leaders. Why it matters: A torrent of data about COVID-19 is being produced, but unless it can be organized in an accessible format, it will do little good. The new initiative aims to use machine learning and human expertise to produce meaningful insights for an unprecedented situation. Driving the news: Members of the newly formed Collective and Augmented Intelligence Against COVID-19 (CAIAC) announced today include the Future Society, a non-profit think tank from the Harvard Kennedy School of Government, as well as the Stanford Institute for Human-Centered Artificial Intelligence and representatives from UN agencies. What they're saying: "With COVID-19 we realized there are tons of data available, but there was little global coordination on how to share it," says Cyrus Hodes, chair of the AI Initiative at the Future Society and a member of the CAIAC steering committee.

CAIAC is the Canadian Artificial Intelligence Association. Formerly known as the Canadian Society for the Computational Studies of Intelligence (or Société canadienne pour l'étude de l'intelligence par ordinateur), CAIAC's mission is to foster excellence and leadership in research, development and education in Canada's artificial intelligence community by facilitating the exchange of knowledge through various media and venues. CAIAC is the official arm of the AAAI in Canada. We are proud to note that Dr. Alan Mackworth, former President of the AAAI, was a founding member of our society. A yearly membership of $30 buys you a significant saving on registration for the annual AI/GI/CRV/IS conference and access to valuable information at this web site such as: Competition Dates and Links, Research Groups and Projects, Recent Doctoral / Masters Theses, Projects looking for Students, Students looking for Projects, Funding Sources, Scholarships and Awards, and Job Opportunities.