Meta remains immersed in the artificial intelligence market with Mark Zuckerberg at the helm. In September last year we were able to witness the birth by Meta AI, the new conversational assistant from the company that directly competed chatbots high-end like ChatGPT from OpenAI. Now, Meta has returned to the AI race with major updates to Meta AI and the release of its latest language model, Meta Llama 3.
This was stated by Meta herself, who announced these new advances in the field of generative artificial intelligence. Meta Llama 3 is proposed as the company’s new next-generation language model, and is also Open source. A model that boasts of having the community as its main pillar, thanks to this open and transparent approach from Zuckerberg.
With these advancements with Meta Llama 3, Meta AI has made new updates and improvements. Zuckerberg’s people announced that this assistant can be used in the food,
Meta Llama 3, the new model
Meta Llama 3 stands out for being a version that presents pre-trained language models, being adjusted to the instructions with parameters 8B and 70B. Meta promises a series of new capabilities that will arrive in this new model in the coming months; Longer pop-ups, additional template sizes will be included and improved performance. However, Llama 3 already benefits from improved reasoning ability compared to the previous generation.
The idea of Llama 3, of course, is open source and an open community approach. With this, Meta intends to make frequent and early releases so that the community has constant access to these models; Actually, It is now possible to try Meta Llama 3 from this link. These new models are the first in a collection of models that will be included in Llama 3.
With this idea, Meta aims to make Llama 3 multilingual, multimodal and able to take advantage of a broader context, with improved performance in aspects such as reasoning or coding. So much so that Llama 3 will be available on all types of platforms, such as AWS, Databricks or Vertex AI; it will be even based on Meta AI, The main Meta assistant who will permeate other departments of the company.
Meta AI will build on Llama 3
Meta AI will be greatly improved with Llama 3, to the point that Meta assures that this assistant will be the smartest assistant available for free to users. To this end, Meta AI will be present on virtually all Meta platforms, including Facebook, Instagram, WhatsApp and Messenger, with English implementation in more than a dozen countries outside the United States.
With this new Meta AI based on Llama 3, it will be possible to do more things faster. For example, receive a new website, meta.ai, it will therefore also be available on computer. The assistant will save conversations with the user as long as they are connected to the platform, being able to access them whenever they want.
Meta AI will be integrated into key aspects of these applications such as search. It will be possible to access information in real time without even having to switch between apps. For example, when using Meta Messenger search, Meta AI can be used to query all question types. The same thing happens with the food from applications like Facebook; he chatbot It will be accessible even from the publications themselves, to offer context or additional information.
However, the key is in image generation, the great advantage of Meta AI. Thanks to the Imagine functionality, Meta AI will offer the possibility of creating images in real time, since when the user starts typing, the generated images will appear in the assistant. These images will also change every few letters written, maintain quality and sharpness at all times.
So much so that Meta AI will provide prompts, with additional ideas for modifying the image, so that the user can continue to iterate from a specific point. You can also animate existing images, with new styles or with the idea of creating GIFs from them. For now, this is a feature reserved for US users of WhatsApp beta and the Meta AI site.