NEW YORK (Reuters) – Metaplatforms on Thursday launched its latest large-scale language model, Llama 3, in real-time while users type prompts, as it races to catch up with generative AI market leader OpenAI. Released an initial version of an image generator that updates images. .
These models will be integrated into Meta AI, a virtual assistant that the company is touting as the most sophisticated of its free-to-use counterpart. The assistant will be given more prominent billing within Meta’s Facebook, Instagram, WhatsApp, and Messenger apps, as well as a new standalone website and more direct interaction with Microsoft-backed OpenAI blockbuster ChatGPT. You will be able to compete.
The announcement comes as Meta looks to challenge OpenAI’s leading position in the technology, including overhauling its computing infrastructure and merging previously separate research and product teams to produce AI products. It comes amid a rush to push the app to billions of users.
The social media giant has equipped Llama 3 with new computer-coding capabilities, this time delivering images as well as text, but for now the model will only output text, Meta chief product officer Chris said.・Mr. Cox said in an interview.
More advanced inference (such as the ability to create longer multi-step plans) will follow in subsequent versions, he added. The version expected to be released in the coming months will also be capable of “multimodality,” meaning it can generate both text and images, Mehta said in a blog post.
“The ultimate goal is to take the hassle out of work and make your life easier, whether it’s dealing with businesses, writing, or planning a trip,” Cox said.
Cox said including images in Rama 3’s training will enhance an update rolling out this year to Ray-Ban Meta smart glasses, a partnership with eyewear manufacturer Essilor Luxoticca, that allows Meta AI to identify objects visible to the wearer. He said he would be able to answer questions. About them.
Meta also announced a new partnership with Alphabet’s Google to include real-time search results in Assistant responses, complementing an existing arrangement with Microsoft’s Bing.
This update expands the Meta AI Assistant to more than 10 markets outside the US, including Australia, Canada, Singapore, Nigeria, and Pakistan. Cox said Meta is “still figuring out the right way to do this in Europe.” Privacy rules are stricter in Europe, with the next AI law expected to impose requirements such as disclosure of model training data.
The voracious need for data for generative AI models has emerged as a major source of tension in technology development.
Meta is offering models like Llama 3 for free commercial use by developers as part of a catch-up effort, as the success of a strong free option could thwart rivals’ plans to monetize proprietary technology. has been released. This strategy has also raised safety concerns from critics who are wary of what unscrupulous developers might build using the model.
Meta CEO Mark Zuckerberg gave a nod to the competition in a video accompanying the announcement, in which he called Meta AI “the most intelligent AI assistant at your disposal.”
Citing metrics used to convey the strength and performance quality of AI models, Zuckerberg said the largest version of Llama 3 is currently trained with 400 billion parameters and has already achieved a score of 85 MMLU. He said there was. He said the two smaller versions currently deployed have 8 billion parameters and 70 billion parameters, with the latter scoring around 82 MMLU (Massive Multitask Language Understanding).
Developers complained that the model in previous Llama 2 versions did not understand basic context, confusing queries about how to “kill” a computer program with requests for instructions to perform a murder. Ta. Rival Google has encountered similar issues, recently suspending the use of its Gemini AI image generation tool after it drew criticism for producing a large number of inaccurate depictions of historical figures.
Meta said Llama 3 has mitigated these issues by using “high quality data” to make the model aware of nuances. The company didn’t go into detail about the datasets it used, but it fed seven times more data into Llama 3 than it used for Llama 2, and used “synthetic” data, or AI, to power areas like coding and inference. He stated that he made use of the data he created.
Cox said there has been “no significant change in attitude” regarding how training data is sourced.
(Reporting by Katie Paul; Editing by Nick Zieminski)
