Jorg Greuel/Getty Images
Post
Post
Share
Annotate
Save
When something goes wrong with an AI system or model, the approaches used to respond to problems caused by traditional software systems don’t work. A new playbook is required. It includes defining AI, identifying relevant harms, designating responders, developing plans to contain the damage in the short term, carrying out a longer-term containment strategy to stop the harm from spreading further, and conducting a post-mortem review.
If there is one constant in the world of technology, it is that the more any given technology is adopted, the more it breaks. This article is about what to do about AI incidents and how to respond to them when they occur.
