BlenderBot 3 was released in the United States last Friday. According to Meta, the program can talk about almost any subject. The chatbot uses artificial intelligence for this, learning from huge amounts of public data that are online.
–
Meta already warned that BlenderBot 3 is a prototype and potentially hurtful comments may appear. When the BBC asked him about Zuckerberg, the bot also said he “did it horribly” when he testified before the US Congress. “I’m concerned that this country is divided and he’s not helping,” the chatbot said. “I’m concerned for all of us.”
–
The company has been criticized for doing too little to tackle the spread of disinformation. Last year, former employee Frances Haugen made headlines after she revealed that Meta believed that making money was more important than users’ online safety.
–
Chatting with the public yields more data
Incidentally, an employee of BuzzFeed a more nuanced picture about Zuckerberg when he asked the chatbot about him. “I don’t have a strong opinion of him,” BlenderBot 3 said. “He’s a good businessman, but his work isn’t always ethical. It’s funny that he has so much money and always wears the same clothes.”
–
Meta will have to swallow these kinds of “witty” answers to collect more data. In a blog post the company writes that the program is being tested in public spaces to chat with regular people. “That leads to longer, more diverse conversations and more feedback that we can work with.”
–
In the past, things often went wrong with public chatbots. Microsoft offered its excuses after Twitter users taught the bot to make racist comments.
–
–