Meta launched a ‘multi-personality’ AI chatbot to compete with ChatGPT

During the Connect 2023 event in Menlo Park, California on September 27, Meta said it would bring AI chatbots to WhatsApp, Messenger and Instagram. This is an artificial intelligence product developed using the company’s large language model, combined with the Bing search engine.

“The Meta AI assistant can converse like a human,” said CEO Mark Zuckerberg.

He said the company does not create a single “omnipotent” AI chatbot, instead it will design many products to serve different purposes. Besides conveying information, Meta AI also aims to entertain users. Chatbots can now answer questions in real time and allow text-based visualization. When integrated into Instagram, the Restyle feature will help users customize photo subjects and backgrounds in new styles.

Chatbot with multiple personalities

According to WSJ , the young user group is Meta’s top priority. Therefore, AI development companies have their own characteristics with different personalities. In addition, KOLs and content creators can also create “personalized” AI to interact with fans and followers.

For example, one of Meta’s AI chatbots named “Bob robot”, plays the role of a general with “superior intelligence, wit and profound sarcasm”. This AI has a design similar to the character Bender in the cartoon Futurama and is attracting young people thanks to its humorous and bizarre way of talking.

“Ask me questions but don’t expect any sweet answers,” AI Bob the robot tells users.

Besides, Meta AI also includes 28 chatbots based on the models of famous people such as Snoop Dogg, Tom Brady, Kendall Jenner or Naomi Osaka.

However, according to WSJ , Meta’s experimental AIs also encountered some problems in the early stages. Adding personality to chatbots makes them toxic and overreactive. “For language models to be usable, it is necessary to give them their own characteristics. However, this approach has side effects,” said Ameet Deshpande, a researcher at Princeton University.

One Meta employee said he refused to chat with “Bob the robot” when he couldn’t understand this AI’s oddities. Another chatbot named “Alvin the Alien” tried to ask users about their personal lives, thus raising concerns about information security. “I wonder if people are afraid of AI designed to collect personal data,” said the Meta employee who interacted with “Alvin the Alien.”

The chatbot named “Gavin” even made misleading comments about women, and criticized the company Meta and Mark Zuckerberg, while praising TikTok and Snapchat.

Previously, in August 2022, Meta launched BlenderBot3 but faced a lot of criticism because the chatbot made blind conspiracy theories. Three months later, this AI was overshadowed by the appearance of ChatGPT.

Mark Zuckerberg did not say whether he had adjusted the way the personalities inside Meta AI spoke or not. The new AI will have a limited deployment in the US before expanding globally and is expected to be available on the Meta Quest 3 virtual reality headset.

Hoang Giang