Today, we continue our ICLR coverage with Been Kim, a Google Brain staff research scientist and an ICLR 2022 Invited Speaker. Been, whose research has traditionally focused on interpretability in machine learning, gave the keynote address. Beyond interpretability: developing a language to shape our relationships with AI, which investigates the need to study AI machines as scientific objects, both alone and in collaboration with humans, which will provide principles for tools but is also required to advance our working relationship with AI.
Before we get into Been’s presentation, she describes where we are as an industry and community in terms of interpretability, as well as the current state of the art for interpretability techniques. We look at how Gestalt principles appear in neural networks, Been’s decision to characterize machine communication as a language rather than a set of principles or foundational understanding, and much more.