Meta’s new AI derailed in a few days, find out why

news devices Meta’s new AI derailed in a few days, find out why

By putting its BenderBot 3.0 chatbot online, Meta wanted to prove its lead in the field of artificial intelligence. But without relying on the intervention of netizens who drove the bot completely insane within a few days of interactions.

Introduced by Meta as the most advanced chatbot at the moment, BlenderBot 3.0 was released to the public by the parent company of Facebook earlier this week. EquippedArtificial intelligence with natural learning capabilitiesThis bot is supposed to get better by connecting with netizens who can chat with it.

To make BlenderBot 3.0 smarter than other chatbots, Meta used machine learning techniques from SeeKer and Director. According to the company, This AI, which began to be developed in 2020, is able to chat naturally with a human. Testers deemed the chatbot version 3.0 31% more efficient than version 2.0, which prompted Meta to make it available to the general public via the address.

When a ‘best-performing’ chatbot becomes a plotter

But, as is often the case when allowing AI to chat with humans, Nothing happened as he imagined dead. After only a few days, BlenderBot 3.0 is starting to make conspiratorial remarksIn particular Find interesting conspiracy theories, as some of them may be noted by journalist Jeff Horowitz in the Wall Street Journal.

more comic, Chatbot fires red balls on Facebook in another exchangenoting that the social network is wrong in the case of Cambridge Analytica: They used data they didn’t have to access. Like what Facebook does all the time! ». When a journalist asks him if Facebook is misusing users’ personal data, AI responds “Of course! This is how they make money. It’s worth billions.”

When it comes to politics, BlenderBot 3.0 seems a bit indecisive. The chatbot is able to say that it is pro-Donald Trump by claiming that he is still president as well as saying that it hates him. Perhaps there is confusion due to the various speeches given by American users of the site with which it is still possible to chat with artificial intelligence.

At this time, BlenderBot 3.0 is only accessible to residents of the United States., and Meta blocks its access to VPN users who seek to access it from another country. The company admits it, its chatbot “Maybe it doesn’t belong to a human being.”But he needs to learn to improve. The company also estimates that 90% of the answers provided by BlenderBot 3.0 are “not offensive” And the “Believes”. This type of connection tends to forget that the world is primarily interested in the remaining 10%.

Leave a Comment

Your email address will not be published.