The Zhitong Finance App learned that Facebook and Instagram parent company Meta Platforms (META.US) have launched the latest upgraded version “Llama 3” of their powerful open source AI model “Llama 3,” and the company also used Llama 3 to update MetaAI, an artificial intelligence chat assistant. This is the latest effort by the social media giant to keep up with similar big model technology from technology companies such as OpenAI and Google (Google) under Alphabet Inc., and the global big model open source community may be able to use GPT-4 AI big models to promote the perfect integration of artificial intelligence technology into all walks of life around the world, thereby comprehensively revolutionizing human social productivity.

Llama 3 released by Meta on Thursday local time is an upgraded version of the Llama 2 open source artificial intelligence model released by Meta last summer. Meta's chief product officer Chris Cox (Chris Cox) said that the big open source model called Llama 2 has been downloaded about 170 billion times. According to information, although Llama 2 is touted as an open source model, its free commercial license has some special restrictions, such as requiring companies with more than 700 million users to separately apply for a special use license, that is, if a company's product has more than 700 million monthly active users (MAU), then they must apply for a special commercial license.

According to information, Meta will open source two Llama 3 8B and 70B models of different sizes for external developers to use for free. In the next few months, Meta will launch a series of new models with capabilities such as multi-modality, multi-language dialogue, and longer context windows. Among them, the strongest version, Llama 3, will have more than 400 billion parameters, which is comparable to GPT-4, which OpenAI is proud of, and Claude 3 launched by Anthropic.

Meta AI chief scientist and Turing Award winner Yann LeCun predicted that more open source big model versions will be launched in the next few months, and indicated that Llama 3 8B and Llama 3 70B are currently the most powerful open source models at the same size, and that the performance of Llama 3 8b is even better than Llama 270 B in some scenarios.

According to reports, Andrej Karpathy, one of the core founding members of ChatGPT developer OpenAI and former Tesla AI director, did not skimp on admiration and praise for Llama 3. As one of the pioneers in the field of big language models, Karpathy believes that the performance of 400B LlamA3 is close to GPT-4, stressing that Llama 3 may be the first GPT-4 level open source big model.

The Meta family of software with 3 billion users will provide the latest version of the AI chat assistant for free

Meta CEO Zuckerberg announced that based on the latest version of the Llama 3 model, the latest version of the AI chat assistant Meta AI provided by Meta, has now been updated and covers Meta family application software such as Instagram, WhatsApp, Facebook, etc., and also has an image generation function, which can generate images based on natural language prompts.

Meta, the world's largest social media company with 3 billion users, is using Llama 3 to upgrade its in-app artificial intelligence chat assistant MetaAI, which also includes various products such as the Ray-Ban smart glasses launched by Meta.

Many large technology companies are currently competing to develop products and services that are deeply tied to artificial intelligence, and are investing billions of dollars in high-tech chips and employee resources to develop AI big language models and other products that focus on artificial intelligence application software. Meta is using artificial intelligence technology to provide new drivers for convenient experiences and video recommendations for its 3 billion users, and is also using AI technology to comprehensively upgrade advertising models and Meta VR headsets.

Meta's chief product officer Cox said Llama 3 is an industry leader in many of its parametric model benchmarks. Cox added that the company is also developing a series of iterations of Llama 3, which will be launched as updated versions.

Meta has been trying to embed artificial intelligence into its products for years, but over the past year, the company's executives have repeatedly emphasized the increased efficiency and user engagement brought about by artificial intelligence in public appearances and interviews. Investors have also always been very optimistic that Meta, which has 3 billion users, can continue to generate larger advertising revenue data with the support of AI technology. Meta's stock price has risen 43% cumulatively since this year.

Meta, which has a huge user base, wants to create a new advertising model! The trend of Meta's stock price reaching new highs over and over again may not stop

Since 2023, the family application software of up to 3 billion users under Facebook's parent company Meta has integrated and deployed dozens of new generative AI functions, covering various mainstream application scenarios such as information search and advertising to commercial information channels. Meta said it is striving to deploy these new generative AI on a global scale. According to analysts, Meta's core business, digital advertising, has resumed growth, and the scale of Meta advertising business is expected to accelerate with the support of new generative AI.

Meta also hopes that more third-party developers will use Meta AI Studio to develop diverse AI chatbots for businesses and creators around the world. These chatbots will be designed for businesses and creators to communicate and interact with users on Meta apps (such as Facebook, Instagram, etc.) on their behalf. Simply put, Meta encourages third parties to develop AI tools and Llama's large model open source resource library to seamlessly connect to the Meta platform and develop AI chatbots that can effectively communicate between different users, thereby comprehensively enhancing the 3 billion user experience and supporting the construction of Meta's commercial and development ecosystem.

As user-centered generative AI applications such as ChatGPT and Sora come out one after another, global companies are participating in the boom in generative AI technology, which may drive a decade-long era of AI prosperity and development. According to a recent report released by Bloomberg industry research analysts, the total revenue scale of the generative AI market is expected to grow from 40 billion US dollars in 2022 to 1.3 trillion US dollars by 2032. This market is expected to double 32 times in 10 years, growing rapidly at a compound rate of up to 42%.

The Bloomberg industry research team said that the market expansion focused first on strong demand for infrastructure needed to train AI systems, followed by strong demand for subsequent terminals using artificial intelligence models, and other service types such as advertising and software applications. The application-side ecosystem for B\ C users, such as advertising and software applications, is undoubtedly the field Amazon, Microsoft, and Meta are most proficient in. Even the AI chip hegemon Nvidia is seeking a share of the pie.

In the digital advertising field that Meta relies on, Meta, which has 3 billion users, strives to fully integrate generative AI into its global advertising business and bring users a new advertising model based on artificial intelligence. This may be an important logic where Meta's recent stock price has repeatedly hit record highs. Wall Street Bank Citi recently raised Meta's 12-month target price from $525 to $590 (as of Wednesday's closing at $501.800). The international bank UBS drastically raised its target price from $530 to $610. Canaccord Genuity is also optimistic that Meta will rise to 610 US dollars recently, drastically raising Meta's target price from 580 US dollars to 610 US dollars.