Title: Training in turmoil: Silent data corruption in systems at scale.

Speaker: Rich Bonderson, Google

Download slides

Abstract: As Silicon process technology marches on and system scale continues increasing, are our traditional methods of ATE and functional testing enough to keep up? What can we do in the face of silent data corruption and quality challenges? Are there new opportunities for efficient solutions for closing test gaps and increasing defect detection and mitigation? We set out with these questions and a call for action in the industry in attacking an increasing problem.

Title: Resilience of AI-driven Applications

Speaker: Ravishankar K. Iyer, University of Illinois Urbana-Champaign (US)

Abstract: The emerging innovations in computer science and engineering are being driven by a data driven AI/ML promise to seamlessly and significantly augment human capabilities. To be successful these systems must interact with other manmade and natural systems with a focus on the flawless capabilities of dynamic decision-making algorithms. Applications, such as, autonomous vehicles, robotic management and control, and enterprise scale computing - all driven by advanced data-analytics and artificial intelligence algorithms, have spawned a level of complexity where traditional methods cannot form the backbone for resilient and safe system design. Using data on operational failures of autonomous vehicles, surgical robots, and extreme-scale computing systems as examples, this talk will discuss features of our research that leverage probabilistic graphical models (PGMs) jointly with deep learning (DNNs) for building and validating a new generation of resilient and safety-critical AI applications.