Meta has announced several new updates to Meta AI on WhatsApp, which is now available in 22 countries and also gains multilingual capability and can understand and respond in languages like Hindi, French, German, Italian, Portuguese, and Spanish.
Starting today, open source is leading the way. Introducing Llama 3.1: Our most capable models yet.
Today we’re releasing a collection of new Llama 3.1 models including our long awaited 405B. These models deliver improved reasoning capabilities, a larger 128K token context… pic.twitter.com/1iKpBJuReD
— AI at Meta (@AIatMeta) July 23, 2024
On Tuesday, Meta announced Llama 3.1, the latest open-source large language model.
Besides the ability to understand and respond in different languages, Meta AI on WhatsApp has also gained new capabilities, including the ability to edit images, where AI can help users remove or replace the background of an image, remove certain objects from the image, and the ability to generate similar AI images.
Unlike some of the AI models, which require an active subscription to access similar capabilities, WhatsApp users can experience these features for free. However, right now, the image edit capability is only available in English, and more languages will follow soon. These experiences are powered by the Llama 3.1 model.
Imagine Yourself is another new feature of the Meta AI on WhatsApp, which allows users to tweak their own images using text prompts like “imagine me with green hair,” “imagine me with tattoos,” “imagine me in an 80s movie,” and more. For now, this feature is limited to WhatsApp users in the US in English and will be expanded to more languages and more countries in the coming days.
Images generated or edited using Meta AI on WhatsApp will be watermarked to help users distinguish between human and AI-generated images.
Besides WhatsApp, Meta AI is also available on Facebook, Instagram, and Messenger. With the latest update, one can access the latest Llama 3.1 model with 405 billion parameters on Meta.ai. Developers can also download the Llama 3.1 model and train it with their own data for free.