In high-stakes scenarios like healthcare– or weeknight “Risk!”– it can be more secure to state “I do not recognize” than to respond to improperly. Medical professionals, video game program candidates, and standard test-takers recognize this, however the majority of expert system applications still like to offer a possibly incorrect solution as opposed to confess unpredictability.
发布者:Dr.Durant,转转请注明出处:https://robotalks.cn/new-method-can-teach-ai-to-admit-uncertainty/