A call for less blame in AI safety incidents draws pushback
One post argues that extreme consternation over safety incidents reduces what the world learns. A reply questions whether a relatively blameless approach is ideal, pointing to incentives to hide harm in other industries.
TLDR
One post calls for sharing safety incidents broadly and relatively blamelessly, treating them as learning opportunities and warning shots. It argues that extreme consternation would counterintuitively reduce what the world learns. A reply challenges that approach to overseeing AI incidents: if it were ideal, wouldn't it already be the norm across industries where people can cause harm and therefore have incentives to hide it?
Combined views
103
1 Source, first seen ago
Comments