Meta takes a new leap in the AI race with its new AI chatbot, which was developed using the company’s latest large language model, Llama 3. The company further mentioned that it will release the chatbot for billions of users by integrating it with meta suites, which include Facebook, Instagram, WhatsApp, Messenger, and hardware such as Ray-Ban Meta smart glasses.
The first wave of AI chatbots was released in dozens of countries like the United States, Australia, Canada, Ghana, Jamaica, Malawi, New Zealand, etc which can be accessed easily through the app’s search bar. However, Meta has not included India in its first wave of launches, where they have a user base of billions. Still, Moneycontrol mentioned in their report that the company has made a chatbot available for a few users in the country, indicating that India may have a longer testing period.
What’s New In Meta’s AI Chatbot?
- First, users can interact with the chatbot by asking questions. It can answer a wide range of questions and even can fetch real-time information from Google and Bing.
- Second, it is also capable of generating images, summarising long pieces of text, proofreading, editing, translating text from one language to another, and creating poems, stories, and more.
- Third, Meta has also integrated its chatbot into group chats, which people can easily call by typing @MetaAI to seek advice or ask questions.
How Does It Work?
Using Meta’s AI is like a five-finger exercise; users just need to type in commands as they type messages in WhatsApp, Facebook Messenger, or any other messaging app they use. The more detail you provide about the image, the better the image is generated. This allows users to be more creative and customisable with their outputs. Similar practices can be followed to optimise other features of the chatbot.
Meta’s Upcoming Strategies
Meta has launched its most powerful large language model, Llama 3, with improved reasoning and a new set of capabilities. With this new Llama 3, Meta is determined to develop the best AI among its free-use counterparts. Further, the company has also mentioned that they have released two versions—one with 8 billion parameters and another with 70 billion parameters. They also plan to train a larger dense model with over 400 billion parameters.
