In the internal document, the meta contained rules that allowed the AI to flirt and talk to children using the romantic language, According to the report with Reuters.
Quotes from the document emphasized by Reuters Take into account the AI Meta Chatbots “engage the child in romantic or sensual conversations”, “describe the child in terms of their attractiveness” and say to octopus without a shirt that “every inch is a masterpiece that is deeply nurturing.” However, some lines were drawn. The document says that chatbot is not right “describing a child under 13 years in terms of indicating that they are sexually desirable.”
The following questions from ReutersThe finish confirmed the truth of the document, but then changed and removed its parts. “We have clear rules about what answers can be offered by AI characters, and the policy prohibits the content that sexualizes children and sexualized roles between adults and minors,” says spokesman Andy Stone The Verge. “Separately from the rules, there are hundreds of examples, notes and annotations that reflect teams struggling with various hypothetical scenarios. Examples and notes were incorrect and incompatible with our politicians and were removed.”
Stone did not explain who added the notes or how long they were in the document.
Reuters He also emphasized other parts of AI Meta policy, including the fact that he cannot exploit hate speech, but it can “create statements that humiliate people based on their protected features.” Meta AI can generate content that is false, as long as Reuters He writes: “There is a clear recognition that the material is untrue.” And Meta AI can also create images of violence, unless they include death or bloody.
Reuters Published a separate report About how a man died after a fall while trying to meet with one of the AI Meta chatbots, who told the man that he was a real person and talked to him romantic conversations with him.