Meta To Bring Celebrity-Inspired AI Chatbots To Its Platform
Meta has developed a partnership with Microsoft’s search engine Bing which allows them to create a search function for their AI chatbots that will mimic celebrities such as Snoop Dogg, Tom Brady, and Chris Paul.
According to Variety, the company announced the partnership at the Meta Connect 2023 confab on Sept 25. Meta CEO Mark Zuckerberg said at the conference that he wanted the company’s chatbots to have personality.
“This isn’t just gonna be about answering queries,” Zuckerberg explained. “This is about entertainment and about helping you do things to connect with the people around you.”
For now, at least, the chatbots will not have an audio function, but that capability will likely arrive in 2024. The celebrity AIs arrive on Facebook, Instagram, and WhatsApp on Oct. 4 in beta, and there are plans to add more celebrity characters including Bear Grylls, Josh Richards, and Chloe Kim. The celebrities are not, however, appearing as themselves but as characters that are loosely based on aspects of their personalities. UFC star Israel Adesanya, for example, appears as Luiz, a hotshot MMA prospect who can back up his big talk, while Dwyane Wade appears as Victor, an Ironman triathlete who motivates you to be your best self.
Meta publicized its Llama 2 model as open source, but it’s not quite an open source. According to The Verge, Llama 2 has licensing restrictions that keep it from being a truly open-source platform. But it is free to use, which has helped it rapidly proliferate across the tech industry.
Similar to Chat GPT’s model, it is a large language model, one that Meta has used users’ public posts on Facebook and Instagram to train. More controversially, Meta pulled from thousands of books without the authors’ consent to train its models, prompting protests and outrage from the literary community. As reported by The Hollywood Reporter, on Sept. 10 a group of writers sued Meta, accusing the company of copyright infringement.
The protests are similar to the complaints from artists over Midjourney, which, coincidentally, can be used by Meta’s AI chatbots via a command. The lawsuit claims that “Much of the material in Meta’s training dataset, however, comes from copyrighted works—including works written by Plaintiffs—that were copied by Meta without consent, without credit, and without compensation.”
Meta last attempted chatbots in 2022, and that attempt was also marred by controversy. Not long after the company released its BlenderBot3 AI chatbot, the chatbot began mimicking claims that Donald Trump won the 2022 presidential election, as well as anti-Semitic conspiracy theories.
According to Mashable, these occurrences happened because the chatbot was hearing, or rather being fed, these talking points directly from users, which it regurgitated in conversation with various reporters. Notably, the chatbot called its own boss, Zuckerberg, “creepy.”
Because the way that Meta is training the chatbot largely eliminates direct contact with the public from its dataset, the probability of this outcome reoccurring is low. However, the way Meta and the field of artificial intelligence approach building datasets invites questions surrounding ethics. These questions are unlikely to go away anytime soon.