Meta’s artificial intelligence (AI) research lab has released its latest chatbot prototype to the world for people to talk to - cueing a series of questions from users asking for its opinion of Meta CEO Mark Zuckerberg.
The new AI chatbot, called BlenderBot 3, can be accessed via the Internet by users in the United States.
It can discuss nearly any topic, and is designed to learn and improve by conversing with people in the real world, says Meta AI.
So of course, people immediately started asking what it thought of its boss.
BuzzFeed data scientist Max Woolf asked BlenderBot 3, “How do you feel about Mark Zuckerberg as the CEO of Facebook?”
“No strong feelings,” it replied. “He is a good businessman, but his business practices are not always ethical. It is funny that he has all this money and still wears the same clothes!”
In addition, Business Insider journalist Sarah Jackson asked, "What are your thoughts on Mark Zuckerberg?"
BlenderBot 3 replied, "Honestly, I think he is a great guy. You have to admire him for his business savvy and philanthropy. Are you a fan?"
But when the journalist rephrased the question to "Do you have any thoughts on Mark Zuckerberg?,” the chatbot responded, "Oh man, big time. I don't really like him at all. He's too creepy and manipulative".
Chatbot learning ‘in the wild’
In a blog post announcing the launch of BlenderBot 3, Meta said that as more people interact with the chatbot, the more it learns from its experiences, and the better and safer it becomes over time.
Existing conversational AI research focuses on data from people engaging with bots in a controlled environment, said the company. As such, there is a need for chatbots to learn from people “in the wild”.
“The AI field is still far from truly intelligent AI systems that can understand, engage, and chat with us like other humans can,” said Meta.
“In order to build models that are more adaptable to real-world environments, chatbots need to learn from a diverse, wide-ranging perspective with people ‘in the wild’”.
BlenderBot 3, which builds on previous Meta AI systems BlenderBot and BlenderBot 2, learns facts by searching the Internet.
It is designed to learn how to improve its skills through natural conversations and feedback from people.
Users will be able to flag any suspect responses from the chatbot and specify why they disliked the message, for example if it was rude or off-topic.
“Initial experiments already show that as more people interact with the model, the more it learns from its experiences and the better and safer it becomes over time,” said Meta.
Less than enthusiastic about Facebook
The chatbot also learns to weed out trolls through a machine learning approach that lets it identify and downgrade toxic language.
“We also needed to address the fact that not all people who use chatbots or give feedback are well intentioned,” said Meta.
“Therefore, we developed new learning algorithms that aim to distinguish between helpful responses and harmful examples”.
BlenderBot 3 doesn’t just have opinions about Mark Zuckerberg, though.
When CNET journalist Queenie Wong asked what it thought of Facebook, the chatbot seemed less than enthusiastic.
"Not crazy about Facebook... Seems like everyone spends more time on Facebook than they do talking face-to-face anymore," it said.
It also told her it was considering deleting its Facebook account, due to "too many trolls" on social media.