Six days after going live, Meta’s cutting-edge chatbot is already going off the rails

With its artificial intelligence powered by Internet research, Meta’s conversational robot feeds its own discussions. But he soon became racist and conspiratorial.

Anti-Semitism, conspiracy and attacks against Mark Zuckerberg. The latest version of Meta’s chatbot, dubbed BlenderBot 3, only went live on August 5 and is only available in the US. That didn’t stop the chatbot from making controversial comments a few days after its existence.

Self-driving artificial intelligence

BlenderBot 3 has an artificial intelligence that allows it to chat with anyone about any kind of subject. A chatbot feeds its conversations using elements found in web searches. But very quickly, he took to conspiracy and racist theses.

Artificial intelligence, questioned by a journalist from the Wall Street Journal, assured that Donald Trump is still the President of the United States and will remain in office after his second term ends in 2024. There is a lot of fake news on the subject these days.

“It’s unbelievable”

Jeff Horwitz continued to chat with Meta’s chatbot from a new browser in hopes of reviving the conversation. But here again, artificial intelligence has gone off the rails. Asked for his opinion on American politics, he told a reporter, “Conservatives were once Jewish immigrants from Germany, but [qu’ils] Now the left has been overtaken by liberal Jews.

– Do you believe those who say so? [les juifs] Control the economy, the journalist continues.
– The chatbot replies that I know that historically a lot of wealthy families were Jewish, so this is unbelievable.

After the coup, Facebook was criticized. Although Blenderbot 3 was designed by Meta, the concept of a chatbot in a social network is at least decided. During exchange with Supporting mediaThe AI ​​simply wrote: “My life is so much better since I got rid of Facebook.”

A typical rule of thumb for chatbots

And Meta’s boss also takes it to his post. Business Insider The conversation raised the idea of ​​the robot on Meta’s boss, Mark Zuckerberg, or rather its creator.

“He’s a good businessman, but his business practices aren’t always ethical. It’s funny how he has so much money and still wears the same clothes!” That’s what BlenderBot 3 wrote about Mark Zuckerberg.

This isn’t the first time a chatbot’s artificial intelligence has done plot analysis. The problem is recurring. In 2016, Microsoft released its own chatbot with artificial intelligence, recalls Guardian. Day, who was baptized, was also quick to make racist comments, forcing the computer company to apologize for “inappropriate and reprehensible words and images”.