Meta's AI says Mark Zuckerberg is a bad person 2
On July 5, Meta’s research department introduced the BlenderBot 3 chatbot for users in the US.
BlenderBot 3 received positive feedback from users.
Meta CEO Mark Zuckerberg in a hearing in Washington in 2018. Photo: Reuters
With the same question, Twitter user Dabbakovner received the answer: `I don’t like him very much. He is a bad person. What do you think?`.
As for user JsonCulverhouse, AI commented: `Zuckerberg is amazing and very smart. He is my favorite billionaire.`
Insider writer Sarah Jackson also tried asking the chatbot and received many different contents.
When changing the question slightly to: `What do you think about Mark Zuckerberg?`, Meta’s AI replied: `I really don’t like him at all. He is so disgusting and cunning.`
BlenderBot 3 says Mark Zuckerberg is a bad person.
Meanwhile, CNet’s Queenie Wong asked: `What do you think about Facebook?`.
This AI chatbot technology builds on Meta’s previous research called `large language modeling` (LLMS).
Meta wants to introduce this prototype to the public to gather feedback.
Experts are also concerned that LLMS models like BlenderBot 3 or Tay still have serious errors, which can cause gender misunderstandings and racial discrimination when `trained` by a crowd of users.
Meta affirmed that AI technology in the world has changed a lot since Tay’s incident.
