Meta Revives 2015 Vibes: Introducing New Chatbots to Its Messaging Apps

Meta is making its move in the AI chatbot arena by introducing a variety of AI-powered bots across its messaging platforms, including WhatsApp, Messenger, and Instagram. As of today, these bots are available in beta to select users in the U.S.

One notable addition is the AI assistant named Meta AI. According to The Verge's hands-on review, it closely resembles popular chatbots such as OpenAI’s ChatGPT and Anthropic’s Claude 2. Meta AI is set to expand to Meta’s recently launched Quest 3 VR headset, as announced by CEO Mark Zuckerberg during the Meta Connect conference. This assistant can facilitate trip planning among friends in group chats, respond to general knowledge inquiries, and browse the internet via Microsoft’s Bing for real-time information.

These new developments remind us of M, Facebook’s previous virtual assistant that aimed to assist with tasks like making plans and managing payments but was only accessible to about 2,000 users in California and ultimately suffered from inadequate support. In contrast, Meta AI is powered by a “custom-made” large language model designed for a wider audience. Unlike M, which required human intervention for certain tasks, Meta AI operates completely autonomously, capable of referencing past interactions and providing “very concise” answers. This could mark a significant improvement.

Meta AI can be activated in any chat. Similar to ChatGPT, it can also generate watermarked images using a text-to-image model called Emu, created by Meta’s AI research team.

Additionally, Meta has unveiled a series of “AI characters” — essentially chatbots crafted to embody specific personalities and emulate celebrities such as Kendall Jenner, Dwyane Wade, MrBeast, Paris Hilton, Charli D’Amelio, and Snoop Dogg. More characters, including Bear Grylls, Chloe Kim, and Josh Richards, are expected to join soon. Each chatbot comes with its own profile image and animates subtly based on the conversation's context.

While the concept is intriguing, recent experiences have shown that even advanced AI chatbots can falter, often fabricating information and missing critical conversation points.

Meta has invested approximately 6,000 hours in identifying potential misuse cases and recruiting staff to engage with the models in preparation for their launch. The company claims to have developed technology to detect and act against content violating its guidelines, restricted non-Meta AI chatbots from web searching, and made its chatbot features accessible to researchers through a bug bounty program.

Only time will tell how these new features perform under real-world conditions.