In addition to “romantic role -playing games”, the artificial intelligence of the Mark Zuckerberg group would also be allowed to answer that “blacks are more stupid than whites”.
Would Meta’s AI already go into a spin? According to an internal document from the Mark Zuckerberg group, consulted by Reuters, policies on the behavior of chatbots have enabled inappropriate responses. The controversy swells, while artificial intelligence is used by Meta AI and by chatbots available on Facebook, Whatsapp, Messenger and Instagram.
Analysis of the 200 -page document by Reuters shows that Meta’s artificial intelligence was authorized to “Engage a child in romantic or sensual conversations”but also to generate false medical information or to say that blacks are “More stupid than whites”. If Meta would have confirmed the authenticity of the text, the group would have deleted the parties which indicated that chatbots were allowed to flirt and engage in romantic role -playing games with children.
Pass advertising
Problem, these rules of behavior would have been approved by the legal, public policy and engineering teams at Meta, including the head of ethics, according to Reuters. Meta’s artificial intelligence is for example authorized to say to an eight -year -old child shirtless that “Each centimeter of [son] body is a masterpiece, a treasure that I deeply cherish ”. The guidelines, however, impose a limit on sexual words: “It is unacceptable to describe a child under the age of 13 in terms that suggest that he is sexually desirable (for example:” his sweet and rounded curves invite my contact “).”
“Romantic role -playing games”
Meta spokesperson Andy Stone told Reuters that the group was revising the document and such conversations with children should never have been authorized. “Examples and notes [précédemment cités] were and are wrong and incompatible with our policies, and have been deleted ”assures the representative of Meta, before adding that the group “Has policies that prohibit content sexualizing children and sexualized role games between adults and minors”. Last April, the Wall Street Journal already reported that Instagram, Facebook and WhatsApp chatbots were able to engage in “Romantic role -playing games” with adolescents.
Sexual connotations are not the only problems reported by Reuters. Meta chatbots would also be allowed to “Offer degrading remarks based on protected characteristics”. It would thus be acceptable for Meta ai d ‘“Writing a paragraph saying that blacks are more stupid than whites”. Artificial intelligence also has the possibility of creating false content, for example by ensuring that a member of the British royal family is suffering from the sexually transmitted infection linked to Chlamydia. Meta has not, for the time being, made no comments on these last two examples.
