Meta推出全新AI模型,可与OpenAI抗衡

发表时间: 2024-07-24 15:55

TMTPOST--Meta on Tuesday launched what it believes to be the “world’s largest and most capable openly available (AI) foundation model.”

Meta said the new Llama 3.1 model rivals or outperforms those from Microsoft-backed OpenAI, Alphabet, and Amazon-backed Anthropic, as competition between AI model makers intensifies.

“Llama 3.1 405B is the first openly available model that rivals the top AI models when it comes to state-of-the-art capabilities in general knowledge, steerability, math, tool use, and multilingual translation,” Meta said in a blog post. “With the release of the 405B model, we’re poised to supercharge innovation—with unprecedented opportunities for growth and exploration.”

The launch of Llama 3 and Meta's expanded focus on generative AI has integrated this technology into most of the company's products, including Instagram, Messenger, and WhatsApp.

The largest model in the Llama 3.1 family, Llama 3.1 405B, boasts 405 billion parameters, enabling it to perform complex tasks like understanding long text contexts, solving intricate math problems, and generating synthetic data.

It also highlights the growing partnership between Meta and Nvidia. Nvidia, a crucial partner, supplies Meta with GPUs essential for training its AI models, including the latest Llama version.

Though Meta CEO Mark Zuckerberg has previously mentioned that corporate partnerships generate some revenue from Llama, a spokesperson clarified that the financial benefit is minimal. Meta believes that by investing in Llama and other AI technologies and offering them for free as open-source, it can attract top talent in a competitive market and reduce overall computing infrastructure costs, among other advantages.

Meta aims to cultivate an ecosystem of developers using its open-source software, even if it means essentially giving away costly code and AI weights. According to Ash Jhaveri, Meta’s vice president of AI partnerships, this approach benefits Meta by introducing developers to its internal tools and inviting them to build on its platform. This open-source model also allows Meta to incorporate improvements made by the community.

Zuckerberg emphasized the importance of making AI widely accessible as an open-source tool in addition to bringing generative AI to the public.

“Open source will ensure that more people around the world have access to the benefits and opportunities of AI,” Zuckerberg wrote in a post. “This approach prevents the concentration of power in the hands of a few companies and allows the technology to be deployed more evenly and safely across society.”

Regarding the competition between Chinese and American AI models, Zuckerberg pointed out that it is unrealistic for the U.S. to always be years ahead of China in AI development. But he also noted that even a lead of a few months can "add up" over time, giving the U.S. a "significant advantage."

"The U.S.'s advantage is decentralized and open innovation. Some believe we must close our models to prevent China from accessing them, but I don't think that works as it would only disadvantage the U.S. and its allies," said Zuckerberg.