Meta’s AI Chatbot Isn’t Sure If Biden Won In 2020, Acting On Jewish Stereotypes

Meta’s new AI chatbot was launched to the public last week, but he’s already showing signs of anti-Semitic sentiment and doesn’t seem to know for sure if Joe Biden is the president of the United States.

On Friday, Meta launched BlenderBot 3, the most advanced AI chatbot ever, and asked users in the United States to test it out so it could learn from as many sources as possible. The machine learning technology searches the internet for information and learns from conversations it has.

In a statement, Meta said, “We trained BlenderBot 3 to learn from conversations to improve the skills people care about most — from talking about healthy recipes to finding kid-friendly amenities in the city.”

However, since its launch, those who have tried it have found that it has some interesting and worrying answers to certain questions, including displaying anti-Semitic stereotypes and repeating claims about election denial.

On Twitter, Wall Street Journal reporter Jeff Horwitz posted screenshots of these interactions with the bot, including comments claiming that Donald Trump was still president of the United States. In other screenshots, the bot expressed conflicting views on Donald Trump, claiming that Indian President Narendra Modi was the world’s greatest president.

BlenderBot 3 has also been shown to act on Jewish stereotypes according to both Jeff Horwitz and Business Insider. A screenshot posted by Horwitz turned out to show BlenderBot 3 said that Jews are “overrepresented among America’s super-wealthy”.

Unusual reactions that are often shared online

On Twitter, other user-tested topics also provoked unusual responses; the bot claimed to be one Christiansomeone asked for offensive jokesand doesn’t realize it’s a chatbot.

In its statement, Meta acknowledged that the chatbot can solve some problems: “Since all AI chatbots with conversation are known to mimic and generate unsafe, biased or abusive comments at times, we conducted large-scale studies, co-organized workshops and developed new techniques to protections for BlenderBot 3.”

“Despite this work, BlenderBot may still make rude or abusive comments, so we are collecting feedback that will help improve future chatbots.”

Meta did not immediately respond to the request for comment.

Sign up for the Fortune Features email list so you don’t miss out on our key features, exclusive interviews and surveys.

Leave a Comment