[ad_1]
At an event in London on Tuesday, Meta confirmed that it plans an initial release of Llama 3 – the next generation of its large language model used to power AI assistants – within the next month.
A report published on Monday by The Information newspaper confirms this It was dead close to launch.
“Within the next month, actually less, and hopefully in a very short period of time, we hope to start rolling out our new range of next-generation foundation models, the Llama 3,” said Nick Clegg, head of global affairs at Meta. Describing what looks like a release of several versions or different versions of the product. “There will be a number of different models with different capabilities and different versatility [released] “Within this year, and it will start very soon.”
Chris Cox, Meta’s chief product officer, added that the plan will be to power multiple products across Meta using Llama 3.
Meta has been playing catch-up with OpenAI, which surprised it and other big tech companies like Google when it launched ChatGPT more than a year ago and the app went viral, turning AI-produced questions and answers into mainstream everyday experiences.
The Meta has largely taken a very cautious approach with AI, but this has not gone down well with the public, as previous versions of Llama have been criticized as being too limited. (Llama 2 was publicly released in July 2023. The first version of Llama was not released to the public, however it was leaked online.)
Llama 3, which is larger in scope than its predecessors, is expected to address this problem, with capabilities to not only answer questions more accurately but also to answer a wider range of questions that may include more controversial topics. She hopes this will make the product attractive to users.
“Our goal over time is to make Meta AI powered by llamas the most useful assistant in the world,” said Joelle Pineau, Vice President of AI Research. “There’s a lot of work left to get there.” The company didn’t talk about the size of the parameters it uses in Llama 3, nor did it provide any demos of how it works. It is expected to contain about 140 billion parameters, compared to 70 billion for the larger Llama 2 model.
In particular, Llama’s Meta families, which are created as open source products, represent a different philosophical approach to how AI should evolve as a broader technology. In doing so, Meta hopes to play a broader role with developers versus more proprietary models.
But Meta also seems to be playing it more cautiously, especially when it comes to other generative AI that goes beyond text generation. The company has not yet launched Emu, its image generation tool, Pino said.
“Response time is very important along with safety along with ease of use, to create images that you are proud of and that represent your creative context, whatever that may be,” Cox said.
Ironically – or perhaps predictably (heh) – even as Meta works to launch Llama 3, it has some significant generative AI skeptics at home.
Yann LeCun, a renowned AI academic who also serves as chief AI scientist at Meta, has criticized the limitations of generative AI in general, and said his bet is on what comes next. He expects that the Joint Predictive Architecture (JEPA), a different approach to training models and generating results, which Meta uses to build more accurate predictive AI will be incorporated into the AI space. Image generation.
“The future of AI is JEPA. It’s not generative AI,” he said. “We’re going to have to change the name of Chris’s product division.”
[ad_2]
Source link
