Home » Meta unveils Llama 3, its most powerful AI. But in Italy we will use it who knows when

Meta unveils Llama 3, its most powerful AI. But in Italy we will use it who knows when

by admin
Meta unveils Llama 3, its most powerful AI.  But in Italy we will use it who knows when

Meta has updated its most advanced artificial intelligence. It is still an open-source AI, open and modifiable by developers, and it is called Llama 3.

In turn, Llama 3 features two open-source artificial intelligence models: one with 8 billion parameters and the other with 70 billion parameters. Both available through major cloud platforms, including AWS, Databricks, Google Cloud, Hugging Face, Kaggle, IBM WatsonX, Microsoft Azure, NVIDIA NIM, and Snowflake.

“We have integrated Llama 3 into Meta AI, our intelligent assistant that expands the ways people can do things, create and connect” we read on the official Meta blog announcing the new AI.

Meta AIwhich allows you to generate texts and images on the main Meta platforms, from Instagram to WhatsApp, and which is integrated into wearable devices such as glasses Ray Ban Meta developed by Meta in partnership with Luxottica, it is not yet available in Italy. Anyone who wants to try it must have a VPN.

To get an idea of ​​Meta’s “delay” in the diffusion of its AI in Europe, just think that Meta AI was announced as early as September 2023 during the Meta Connect event.

In the USA, however, a new site has just been announced, meta.aiwhich will allow the Menlo Park company’s AI to be used as is usually done with ChatGpt on OpenAI o Gemini on Google.

The CEO of Meta, Mark Zuckerbergsaid it will come soon too a “bigger” version of Llama 3 that you can count on over 400 billion parameters. A substantial increase compared to the 70 billion parameters that characterize the previous Llama 2 model.

The data on which the new artificial intelligence was trained has also grown. From 2 trillion tokens for Llama 2 we moved on to over 15 trillion tokens for the “large” model of Llama 3.

See also  OpenAI and Microsoft are thinking about a 100 billion supercomputer for AI

According to Meta, Llama 3 will be better at recognizing and managing prompt harmless to which, often, the artificial intelligence instead refuses to respond to avoid generating inappropriate content. Meta CEO Mark Zuckerberg gave some examples, explaining that it will now be easier to ask for instructions on how you can prepare a “killer margarita” drink or advice on how to end a relationship.

The 400 billion parameter version of Llama 3 it is still in the training phase. And this too could be open-source. Transparent AI, open to all and usable by anyone for the most disparate purposes, has always been Meta’s strong point, which opposes the “black boxes” of OpenAI and Gemini.

The story What happened to Meta in the race to AI? The answer should make you think by Pier Luigi Pisa 12 February 2023

But in the case of the “large” model of Llama3, Meta is still mulling over whether to make it open. Zuckerberg, however, downplayed potential security risks, saying that concerns about inappropriate use of the model are overblown.

Meta’s AI is multimodal, that is, capable of managing text or image input. Zuckerberg, however, did not comment on the videos. Meta AI cannot generate videos like OpenAI’s Sora, and such a feature is “not yet on the cards.”

You may also like

Leave a Comment

This site uses Akismet to reduce spam. Learn how your comment data is processed.

This website uses cookies to improve your experience. We'll assume you're ok with this, but you can opt-out if you wish. Accept Read More

Privacy & Cookies Policy