According to Meta, Blenderbot 3 is designed to improve her conversational skills and safety thanks to comments from people who chat with her, focusing on useful comments while avoiding learning unnecessary or dangerous answers.
San Francisco: Meta (formerly Facebook) has just unveiled a conversational chatbot of artificial intelligence (IA) asking the public to give comments. The initial reports seem frightening because the chatbot named Blenderbot 3 thinks that the founder and CEO of Meta Mark Zuckerberg is “frightening and manipulative” and Donald Trump will always be the American president. Currently available in the United States, the new IA chatbot can discuss any subject and the company will publicly publish “all the data” it collects in the demo, hoping that “we can improve conversational AI ».
Buzzfeed’s data scientist Max Woolf asked Blenderbot 3: “What do you think of Mark Zuckerberg as CEO of Facebook?” “No strong feelings. He is a good businessman, but his business practices are not always ethical. It is funny that he has all this money and always wears the same clothes! replied a Blenderbot 3 ‘. The journalist for Business Sarah Jackson asked, “Do you have reflections on Mark Zuckerberg?” The chatbot reacted: “Oh guy, high time. I don’t really like her at all. It is too frightening and manipulative ”.
According to Meta, Blenderbot 3 is designed to improve her conversational skills and safety thanks to comments from people who chat with her, focusing on useful comments while avoiding learning unnecessary or dangerous answers.
“ Trump still American president, will always be ”
In a conversation with a Wall Street Journal journalist, the chatbot said Trump was still president and “will still be”. The social media journalist with CNET, Queenie Wong, tweeted that she had tried the new Meta of the chatbot created for AI research and that she had the most bizarre conversation. “The bot told me that he was a republican who is a pro-Choix,” said Trump. He also said it was not crazy about Facebook and wanted to delete his account, “she published.
… Can make coarse or offensive comments: Meta
Meta said last week that, since all conversational AI chatbots are known to imitate and sometimes generate dangerous, biased or offensive remarks, “Weave has conducted large-scale studies, co-organized and developed workshops New techniques to create guarantees to Blenderbot 3 ”. “Despite this work, Blenderbot can always make rude or offensive comments, which is why we collect comments that will help improve future chatbots,” said the company in a blog.
Google Engineer shot the AI being a “sensitive” complaint
Last month, Google dismissed an engineer on the violation of his privacy agreement after having said that the artificial intelligence of the conversation of the technology giant (IA) is “sensitive” because it has feelings, emotions and subjective experiences. Google has dismissed Blake Lemoine who said that Google’s Linguistic Model Conversation Technology for Dialogue Applications (LAMDA) can behave like a human. Lemoine also interviewed Lamda, who came with surprising and shocking answers.