Daniel Filan Presents “Exploring Hierarchy-Aware Inverse Reinforcement Learning”

13 Jul 2018

CHAI’s Daniel Filan traveled to Stockholm, Sweden to present this paper at the 1st Workshop on Goal Specifications for Reinforcement Learning, FAIM 2018. The abstract reads:

CHAI’s Anca Dragan and Smitha Milli Publish Paper on “Model Reconstruction from Model Explanations”

Two researchers from CHAI, Anca Dragan and Smitha Milli, published this paper together in arXiv along with a few other researchers from Berkeley. The abstract reads:

Thomas Krendl Gilbert Publishes “A Broader View on Bias in Automated Decision-Making – Reflecting on Epistemology and Dynamics”

02 Jul 2018

CHAI’s Thomas Gilbert and his colleagues at UC Berkeley recently published this paper to arXiv and presented it at the 2018 International Conference on Machine Learning. The abstract reads:

Daniel Filan Posts on Mechanistic Transparency for Machine Learning in the AI Alignment Forum

10 Jun 2018

CHAI’s Daniel Filan’s blog post Mechanistic Transparency for Machine Learning posted on the AI Alignment Forum. You can read the blog post here.

How the Enlightenment Ends

01 Jun 2018

Henry Kissenger writes about the rise of artificial intelligence. Read the article here.

Anca Dragan Publishes “Probabilistically Safe Robot Planning with Confidence-Based Human Predictions”

31 May 2018

The paper published in arXiv and supported by the National Science Foundation deals with human-computer interaction and having robots be able to navigate around moving people properly. The abstract reads:

CHAI’s Rohin Shah Creates New Newsletter for AI Alignment

23 Apr 2018

CHAI PhD student Rohin Shah created this new newsletter for everyone involved in AI Safety to keep up to date on research in the field.

Joseph Halpern Publishes “Towards Formal Definitions of Blameworthiness, Intention, and Moral Responsibility”

05 Apr 2018

CHAI’s Joseph Halpern and his colleague Max Kleiman-Weiner at Cornell University published this paper in the Proceedings of the Thirty-Second AAAI Conference on Artificial Intelligence (AAAI-18). The abstract reads:

Anca Dragan – Helping Robots Express Themselves When They Fail

27 Mar 2018

Anca Dragan’s research was featured in this article by the IEEE on how robots can communicate that they can’t accomplish a certain task. You can read the article on IEEE’s blog here.

Anca Dragan Teaches Robots How to Behave

08 Mar 2018

Anca Dragan and Andrea Bajcsy from UC Berkeley, along with two other researchers from Rice University, published a paper arXiv called “Learning from Physical Human Corrections, One Feature at a Time” The abstract reads: