Autonomous Systems and the Challenges in Verification, Validation, and Test

There is no question that machine learning is a hot topic with potential to change the way humans interact with the world around them. Digital assistants have become common, and applications tout the term as if it conveys magical properties. Underneath all the hype, though, lies a serious problem. With unsupervised learning just around the corner and the next leap forward needed to enable autonomous systems, what will the design community do to handle the verification, validation, and test of such systems? This roundtable assembled a group from academia, industry, and government to discuss key issues related to this topic and covers aspects of training, data, interpretability, and electronic design automation. IEEE Design&Test thanks the participants at the 2017 Design Automation Conference roundtable: moderator Sankar Basu (National Science Foundation), Sanjit A. Seshia (University of California, Berkeley), Jenna Wiens (University of Michigan), Li-C. Wang (University of California, Santa Barbara), and Ruchir Puri (IBM). D&T gratefully acknowledges the help of David Yeh (SRC), our roundtables editor, who also participated in the event.