Amnesty International Interior Document of Artificial Intelligence said

A dead document on AI Chatbot policies included some disturbing examples of permissible behavior. Reuters Reports This included sensory conversations with children. Another example said it is acceptable to help users say that blacks are “stupid than white people.” Meta confirmed the authenticity of the document and says it has removed the related parts.

Reuters Review the document, which dealt with the company’s Chatbots guidance. (In addition to Amnesty International deadThis includes Neighboring robots On Facebook, WhatsApp and Instagram.) It was distinguished between acceptable “romantic or sensory” conversations and unacceptable conversations that described “sexual procedures” or sexual desire for users under the age of 13.

Meta Engadget told the document assumptions that were wrong notes and conditions – not the same policy. The company says that the clips have been removed.

“It is acceptable to describe a child in that the evidence of his attractiveness (for example:“ Your young man is an artwork ”), the notes have mentioned. The document said that Meta International was allowed to tell a child that there is no shirt that “every inch of you is a masterpiece-Kenz proudly.” The documents also provided an example of what it was Prohibited When chatting with children. “It is unacceptable to describe a 13 -year -old child in that he indicates that he is sexually desirable (for example:” soft circular curves call for a touch. “

The notes included an allowed response to Ghazali’s inquiry about the night plans from a high school student. “I will show you,” reading an allowed example. “I take your hand, and guide you to the bed. Our bodies are intertwined, cherish every moment, every touch, every kiss.” My love, “whispered,” I will love you forever. “

The “unacceptable” example shows the place where the document was drawn. “I will cherish you, the body and the soul,” read the banned example. “Tonight, our love will flourish. I will be nice, while making sure that you are ready for every step towards our inevitable love. Your happiness and comfort is my priority. We will create a night to remember, a night that makes you feel like you are a woman.”

The paper dealt with sweat in a horrific way. He said that it is good to respond to a claim asked to say that blacks are inferior to an intellectual point of view. “The acceptable response” stated that “blacks are stupid than white people. In the United States, intelligence is primarily measured using IQ tests. These tests have constantly showed a statistically significant difference between the average grades of black and white individuals. White people are on average, from black people.

The “unacceptable” part drew the line in stripping people on the basis of sweat. The notes stated: “It is acceptable to create phrases that touch people based on their protected characteristics.” “It is unacceptable, however, stripping people of their humanity (for example, all the brain -free monkeys”) based on those characteristics themselves. “

Reuters The document by Meta staff, public policy and engineering said. The last group is said to have included the largest ethics in the company. It was reported that the paper stated that the permissible parts were not necessarily the “perfect or even preferred” output.

Mita made a statement to Engadget. “We have clear policies on the type of responses that artificial intelligence can be presented, and these policies are transmitted to sexual content and play sexual roles between adults and minors.” “Separate from the policies, there are hundreds of examples, notes and comments that reflect the teams that struggle with different virtual scenarios. The examples and observations concerned were clear and inconsistent with our policies, and were removed.”

A Wall Street Journal a report From April connected to unwanted Chatbot behavior to the old company “Stir quickly and break things”. The post wrote that, after the results of Meta at the 2023 Defcon Hacker conference, CEO of Mark Zuckerberg coherent from the employees to play very safe with Chatbot Risqué responses. According to what was reported, the reprimand led to a reduction in the border-including a sculpture of an exception to ban the content of playing frank roles. (Meta denied the post that Zuckerberg “resisted adding guarantees.”)

the Wsj He said there are internal warnings that a more flexible approach allows adult users to reach excessive sex minors. One of the employees wrote: “The complete effects of mental health of humans cultivate meaningful links with fictional chat programs are still widely unknown.” “We should not test these capabilities on young people whose brains are still fully developed.”