This and all episodes at: https://aiandyou.net/ .
Training an AI to render accurate decisions for important questions can be useless and dangerous if it cannot tell you why it made those decisions. Enter explainability, a term so new that it isn't in spellcheckers but is critical to the successful future of AI in critical applications.
Before I talked with Michael Hind, my usual remark on the subject was, "If you want a demonstration of the ultimate futility of explainability, try asking your kid how the vase got broken." But after this episode I've learned more than I thought possible about how we can teach AI what an explanation is and how to produce one.
Michael is a Distinguished Research Staff Member in the IBM
All this plus our usual look at today's AI headlines.
Transcript and URLs referenced at HumanCusp Blog.