In several previous posts here, I've been discussing the risks of biased AI, particularly in natural language processing tasks. It's part of my work at Luminoso to address this, both in ConceptNet and in Luminoso's services. We're not just warning about what not to do, we're promoting best practices about what you can do to prevent bias right now.
At Luminoso, Denise Christie and I recorded a discussion about de-biasing, touching on a lot of the current issues that create bias in NLP and what to do about them. The recording and transcript are now available on Luminoso's site.