32Parents want an alert when their teenager's AI chats show signs of distress. Teens want privacy. Design the alert.▼expertNewOpenAIMeta◆ premiumAt a low base rate even a good classifier mostly flags teens who are fine. Decide what triggers an alert, who reviews it, what the parent sees and what the teen is told.Practise this question →Read answer →
38Our general-purpose assistant handles 100 million messages a week. Design how it responds when a user may be in crisis.▼hardNewOpenAIAnthropic◆ premiumAt a 0.05% base rate, a classifier that is right 99% of the time on safe messages still flags about twenty people who are fine for every one in danger. Build a ladder of responses that each fit their error rate.Practise this question →Read answer →