news

한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina

July 24 news, local time in the United States on Tuesday, Facebook's parent companyMetaIntroducing a powerful newAImodel, which was named by Meta CEOmark Zuckerberg(Mark Zuckerberg) called it "frontier-level" and aims to work with OpenAI andGoogleCompete with similar products from industry leaders such as

Meta spent months training the new AI model, called Llama 3.1, and spent hundreds of millions of dollars to improve computing power. The company said it was a major upgrade to Llama 3, which was released earlier this year. Zuckerberg emphasized in the interview: "In the field of AI assistants, intelligence is the core of product competitiveness. The Llama model we are building will be one of the most advanced AI models in the world." He also revealed that Meta has started to develop Llama 4.

Meta executives said that this new model is designed to drive chatbots for Meta's internal and external developers, and has many new features, including enhanced reasoning capabilities that can assist in solving complex math problems or instantly generate the text of an entire book. In addition, Llama 3.1 integrates multimodal technology, allowing users to instantly generate images through text prompts. Particularly worth mentioning is the "Imagine Yourself" feature, which allows users to upload personal facial images and then generate personalized depictions in a variety of scenarios.

Through the Llama model, Meta powers its artificial intelligence chatbot called Meta AI. This chatbot is not only seamlessly integrated in its core applications such as Instagram and WhatsApp, but is also popular as a standalone web product. Zuckerberg said that Meta's chatbot has "hundreds of millions" of users and predicted that by the end of the year, it will become the most widely used chatbot in the world.

Zuckerberg also expects other companies besides Meta to use Llama to train their own AI models. He said: "The Llama model will serve as a building block for many organizations to create their own models instead of having to rely on ready-made models on the market."

Meta's investment in artificial intelligence is huge. Zuckerberg admitted that the training of the Llama 3 model alone cost "hundreds of millions of dollars" and expected the cost of developing future models to climb to the "billions of dollars" level. Although Meta cut some future technology and management expenses and cut thousands of jobs in 2023 in the "Year of Efficiency", Zuckerberg is still willing to invest heavily in the artificial intelligence competition.

Zuckerberg pointed out: "Although many companies now seem to be overinvesting, looking back in the future may feel that we have spent billions of dollars unnecessarily. But on the other hand, all companies that make investments are based on reasonable decisions, because lagging behind means being at a disadvantage in the key technology competition in the next ten to fifteen years."

After making a large investment, Meta decided to open Llama technology for free use by the public, provided that users abide by the company's "Acceptable Use Policy." Zuckerberg hopes that this open access strategy will help Meta's work become the basis for other successful startups and products, thereby exerting a greater influence in the industry.

"If AI is going to be as important as mobile platforms in the future, then I don't want to be in a position where I can get AI from a competitor," he added. Zuckerberg has long expressed dissatisfaction with Meta's reliance on Google and Apple's phones and operating systems to release social media apps. "We're a technology company and we need to be able to build things ourselves at every level. So it's worth making these large investments," he said.

Zuckerberg debunked rumors about the advantages of using Facebook and Instagram post data to train Llama models, noting: "A lot of public data is already indexed by search engines, so I think companies like Google can use that data, too."

In April, Meta told investors that its annual spending was expected to exceed its initial expectations by several billion dollars, thanks to investments in artificial intelligence. According to a company blog post, Meta expects to have about 350,000 Nvidia H100 GPUs by the end of the year. These chips are the basis for training large language models such as Llama and OpenAI's ChatGPT, and each chip can cost tens of thousands of dollars.

Although Meta's open source AI strategy has been criticized, with major concerns including the risk of abuse and the use of technology by geopolitical rivals such as China, Zuckerberg is more concerned about the long-term damage that closed technology may cause. He believes: "It is wrong to close all technologies because the prosperity of the United States is built on open and decentralized innovation. Closed technology will only hinder our development and make us less likely to become a leader."

He also pointed out that while it is unrealistic for the United States to maintain its lead over China in the field of artificial intelligence for many years, even a short-term lead can accumulate over time and bring a clear competitive advantage to the United States.