Remember when we told you at the end of last week that Meta had just launched a universal chatbot? I’ll admit that I was a bit pensive after posting that news. Maybe he was too critical? Perhaps the Meta developers have considered the risks and planned to surprise us by making BlenderBot immune to the distortion attempts that can be expected from the start.

At the end, We are talking about very intelligent people, right? Otherwise they wouldn’t be where they are. So, I’ll admit, I tried to access BlenderBot a few times this weekend to do some testing and check if I was over my pessimism. Unfortunately, all attempts were unsuccessful, so I’ve been dragging my feet ever since until a few minutes ago.

And I’m sorry to say it, but to no one’s surprise, it happened again. As we can read in Bloomberg, 72 hours was enough for the Meta chatbot to look like a troll on Twitter. Moreover, he is not even consistent in his approaches. And so it is that while some users read a bot that stated that Donald Trump won the 2020 election and that everything that happened after that was a fraud, while in other conversations BlenderBot would acknowledge the legitimacy of said election and the presidency of Joe Biden. To complete the game, a third person claims that BlenderBot expressed explicit support for Bernie Sanders in his interview.

On the other hand, the chatbot also subscribed to some conspiracy theorysuch as the overrepresentation of Jews in the economic elite of the United States, pointing to the existence of their plan to dominate the global economy, which is a very common theory in anti-Semitic circles and that, in the case of the Meta chatbot, has already been condemned by the League Against the Defamation of the Jewish People.

And, of course, we’re talking about Meta, Mark Zuckerberg’s company, something the trolls weren’t going to let go of, more than imaginable. At the request of the creator of the social network described Meta CEO Mark Zuckerberg as “too creepy and manipulative«.

Meta acknowledges that their chatbot can say offensive things because it’s still an experiment in development and users actually have to check a box when accessing the chatbot that says: “I understand that this bot is for research and entertainment purposes only and may make false or offensive statements. If this occurs, I am committed to reporting these issues to help improve future research. I further agree not to intentionally trigger a bot to make offensive statements.“.