Meta Platforms has released an early version of its latest major AI language model, called Llama 3, as well as an image generation tool that updates them in real-time as the user inputs queries. As Reuters writes, this move shows Meta's desire to catch up with the market leader in generative AI, OpenAI.
The new models integrate the Meta AI virtual assistant, which the company calls the best among free counterparts, citing better results compared to competitors such as Google (from Alphabet) and French startup Mistral AI. The evaluation took into account the ability to think logically, write code and creative writing.
The updated Meta AI will get a more prominent place in the Facebook, Instagram, WhatsApp and Messenger apps, and will also have a separate website. He will be able to perform various tasks, for example, create a list of things for the trip, help with homework or write creative texts.
Llama 3 was taught not only to understand text, but also images. This will improve the virtual reality glasses Ray-Ban Meta, which will be able to answer the user's questions about the objects he sees.
Meta also announced a partnership with Google to include real-time search results in Assistant responses, complementing its existing agreement with Microsoft's Bing search engine.
In addition, the Meta AI Assistant is now available in more than a dozen markets outside the US, including Australia, Canada, Singapore, Nigeria and Pakistan.
Meta also tries to fix problems that were in the previous version of Llama, such as the inability to understand context. The company says it used high-quality data to train Llama 3 so the model can recognize nuances.