Wednesday, Meta declared Its consumer-friendly entry into the crowded AI chatbot landscape, on the edge reports. During a presentation at Meta Connect 2023, the company said it will introduce its own “Meta AI” chat assistant and a selection of AI characters on its messaging platforms, including WhatsApp, Instagram and Messenger.
Meta’s new AI assistant will be familiar to anyone who has used chatbots like ChatGPT or Claude. It’s designed as a general-purpose chatbot that Meta says can help plan trips, answer questions, and create images from text prompts. Assistant integrates real-time results from Microsoft’s Bing search engine and provides access to current information such as Bing Chat, ChatGPT’s browsing plugin, and Google Bard.
During demos, Meta’s AI was able to quickly generate high-resolution images from short text descriptions using the “/imagine” prompt, and the feature will be free to use, according to The Verge. While Meta didn’t reveal the full details of the new AI assistant’s training, the company said the model is a custom model partially based on the company’s LAMA 2 language model released in July.
In addition to the general assistant, Meta is releasing 28 AI-powered characters on its messaging platforms, including Tom Brady (a sports fan who plays “Bru”), Kendall Jenner (a host of celebrities who play “Billy,” “big sister”), and Snoop Dogg (“Dungeon Master” with gaming advice).
According to TechCrunch, celebrity images aren’t quite videos, but can be created to function as innovative animations. However they are said not to speak, so they are more like representative avatars of text-based AI bot personalities. Meta says the popular AIs went into limited beta on Wednesday, but we’ve yet to see reports of them on the ground (or on the phone) in action.
The company says it spent 6,000 hours on red team exercises to identify and resolve potential problematic uses of its assistants to ensure these popular AI characters don’t cross into Tay territory and spew embarrassing or harmful phrases. This process involves creating thousands of internal conversations to refine the assistant’s behavior and responses.
For now, Meta’s new assistant isn’t trained on public user data from Instagram or Facebook. However, The Verge points out that this could change as the company aims to improve the usefulness of its chatbot through “social integrations,” according to Ahmed Al-Dahle, Meta’s VP of Generative AI.
Meta executives see the company’s massive user base as a key advantage, with billions of daily users across WhatsApp, Instagram and Messenger. While OpenAI arguably pioneered the creation of the AI chatbot space, Meta’s scale will allow its assistant to reach more consumers for their first chatbot experience — while driving profitable user engagement with its apps.