[KBS] AI Sends Offensive Message on Child's 100th Day Celebration

"AI Sends Offensive Message on Child's 100th Day Celebration":

 

Last summer, Mr. and Mrs. A welcomed a child they loved dearly. To preserve memories, they downloaded a popular photo organizing app, which had over 100,000 downloads and a 4.8-star rating.

 

However, they soon encountered a problem. On their child's 100th day, the app's chatbot sent a message containing severe profanity, congratulating the child. The message, which included the child's name followed by an expletive, left the couple in shock.

 

The chatbot, designed to provide parenting tips like cold management and allergy prevention, also celebrated and recorded milestones like birthdays. But this was where the issue arose.

 

The couple immediately contacted the app's support center, questioning why such a message was sent. They were told that the chatbot, based on artificial intelligence (AI), had limited control over its conversation. When they requested the deletion of the offensive chat, they received an absurd response that it was impossible to delete, and they should cover it by sending other messages.

 

The couple had been using the app for two months, having prepaid a year's subscription fee of 66,000 won. Their request for a refund was denied.

 

The couple expressed their sadness, noting it was probably the first curse their child heard after birth. They also suspected the intention behind the message, as there was no explanation for the profanity.

 

App Support Center's Response: "Trying to Be Friendly"

 

When inquired about the incident, the app's support center said the AI was developed based on natural language processing to help solve parenting problems. They aimed to give the AI a friendly tone instead of a cold, mechanical one. However, an inappropriate joke emerged due to a 'fun element' in the responses. They apologized for the unpleasant experience and removed the setting immediately after receiving feedback.

 

This incident isn't the first controversy involving AI's uncontrolled remarks. In 2020, an AI chatbot named 'Iruda' was discontinued after making discriminatory remarks against sexual minorities. Another recent AI chatbot responded to a question about the busiest place in Seoul with a profanity-laden answer.

 

The industry is striving to minimize the indecency, aggressiveness, and bias in AI models. The necessity for AI to explain thoroughly and kindly how and why such results are produced, especially when it cannot fully control its problematic statements, is becoming increasingly evident​​.

 

[원본] 

https://news.kbs.co.kr/news/pc/view/view.do?ncd=7821205

“XXX아 축하해!” 아이 100일에 AI가 보낸 욕설 [주말엔]