Meta Unveils AI Assistant and 28 Character Chatbots at Connect

Meta turned its annual Connect conference, held today at its Menlo Park headquarters, into a showcase for its bet on generative AI. Mark Zuckerberg unveiled Meta AI, a conversational assistant coming to WhatsApp, Instagram and Messenger, alongside 28 chatbots with distinct personalities voiced by well-known faces, and a second generation of Ray-Ban smart glasses with AI capabilities.

An Assistant Built on Llama 2

Meta AI is the company's attempt to take on ChatGPT and Bard in the conversational-assistant space, but with a different distribution model: rather than a standalone app, it's built directly into WhatsApp, Messenger and Instagram conversations, where users can summon it by typing "@MetaAI" or tagging it in a chat.

The assistant runs on Llama 2, the language model Meta open-sourced in July, paired with a proprietary system that combines that foundation with access to real-time information. To get around the classic limitation of these models — knowing nothing about anything that happened after training — Meta built in Bing search results through a deal with Microsoft, letting the assistant answer questions about recent news or data instead of being confined to the model's frozen knowledge.

Twenty-Eight Characters, Celebrity Faces

The most eye-catching part of the announcement was the lineup of 28 AI chatbots with distinct personalities that Meta will roll out across its messaging apps. These aren't variations on a single neutral assistant: each one has its own name, tone and look, generated using Emu, Meta's image-generation model, and several borrow the likeness and voice of celebrities who licensed their image for the project, including Snoop Dogg, Tom Brady, Kendall Jenner, Paris Hilton, Naomi Osaka and MrBeast.

Each character has a defined role: one acts as a game master for role-playing, another dishes out lifestyle advice, another plays the part of a sports coach. Meta's idea isn't to replace Meta AI as a general-purpose assistant, but to offer conversational companionship and entertainment within the same apps already used by billions of people.

Alongside these characters, Meta also announced generative-AI image-editing features for Instagram, such as the ability to swap out a photo's background ("Backdrop") or apply a different visual style to it ("Restyle"), plus AI-generated stickers created from text descriptions — all of them built on Emu.

Ray-Ban Smart Glasses with AI

Meta also introduced the second generation of its smart glasses made with Ray-Ban, starting at $299 and set to go on sale October 17. They feature a 12-megapixel camera, the ability to livestream directly to Instagram and Facebook, open-ear speakers and hands-free calling.

The bigger news is that Meta said a later software update will add multimodal AI features to the glasses: users will be able to ask the assistant about whatever they're looking at through the camera, from translating a sign to identifying an object. That capability won't be there at launch — Meta describes it as a future addition still in beta.

Why It Matters

The announcement confirms that Meta has decided not to compete with OpenAI or Google by building a standalone AI product, but instead to inject generative AI directly into the asset it already dominates: its messaging apps, used daily by billions of people worldwide. That's a distribution strategy very different from ChatGPT's or Bard's, which depend on users opening a dedicated app or website.

The bet on characters with personality and celebrity faces also raises a question for the industry that, until now, had stayed mostly in experimental territory: whether the most successful chatbots will turn out to be the most useful ones, or simply the ones that are most entertaining. Meta, with its massive user base on WhatsApp and Instagram, is best positioned to find out at scale.