Exploring the Path Towards Safe and Robust ML in Real-World Systems

Modern ML algorithms are opaque, fragile, and can be susceptible to data poisoning and tampering, as well as a variety of perturbations. The shift towards deployment of AI in real-world systems demands that we assess whether our ML algorithms are safe, robust, and secure. Join us as we consider how these problems may drastically and unexpectedly impact the performance and behavior of our systems.