On January 18, 2024 local time in the United States, Meta CEO Mark Zuckerberg announced a large-scale restructuring of the AI department and began training the large model Llama 3. Although the metaverse has already cost Meta billions of dollars, Zuckerberg has no intention of giving up on the metaverse. He stated that Meta is expanding the size of its AI team and is committed to promoting the development of the metaverse and artificial intelligence.
Zhang Yi, CEO of iMedia Consulting, told China Business News that Zuckerberg's adjustment is actually very obvious. Firstly, Zuckerberg is optimistic about the future prospects of the metaverse. However, helplessly, this business cannot generate direct economic benefits, as reflected in financial data. With only inputs, there is no better output, and the capital market is not entirely willing to pay for it. In addition, the AI big model has a mature business model, and its competitor, Open AI, has received huge returns as a result. Of course, for Meta, in the past year, it has gained a certain degree of revenue from launching AI big model products and also gained recognition from the capital market.
Hoarding computing power
On January 18th, Zuckerberg posted a video on social media. In the video, Zuckerberg stated that it is becoming increasingly clear that the next generation of services requires building comprehensive universal artificial intelligence, building the best AI assistance, AI for creators, AI for enterprises, and so on. This requires progress in various fields of AI, from reasoning to planning, from coding to memory and other cognitive abilities.
Talking about the application scenarios of Meta's AI, During an interview with our reporter, Zhang Fa'en, CTO of Innovation Intelligence, stated: Meta is a leading social software company in the world, just like WeChat in China. Overall, their application scenarios are very diverse. I think Zuckerberg's AI work is considered from a fundamental perspective. Currently, Meta has invested a lot in AI, including the current LLama series, where Llama1 and Llama2 have been launched, and Llama3 is about to be released. These are the research and development results of very common AI technologies, which have a wide range of application scenarios in social media For example, automatic chatbots. If a person has emotional disorders, they can interact with Meta's emotional teacher, which is supported by a chat application provided by a large model. I think there will be a lot of application scenarios like this
In his video post, Zuckerberg emphasized Open Source. He stated that Meta's long-term vision is to establish universal artificial intelligence, responsibly open source it, and make it widely available for everyone to benefit from. Currently, Meta is closely linking the two main artificial intelligence research groups, FAIR and GenAI. Zuckerberg also revealed, "We are currently training the next generation model Llama 3." To this end, Meta will prepare to gather the most powerful AI computing power in the industry. Zuckerberg stated that the company will purchase over 350000 Nvidia H100GPUs - currently the most powerful chip in the industry for building generative AI performance.
According to research from third-party investment institutions, Nvidia's H100 shipment to Meta is estimated to reach 150000 yuan in 2023, which is on par with its shipment to Microsoft and at least three times that of other companies. Zuckerberg stated that if Nvidia A100 and other artificial intelligence chips are included, Meta's GPU computing power will reach an equivalent of nearly 600000 H100 by the end of 2024.
When it comes to why Zuckerberg wants to do metaverse and AI, Zhang Fa'en believes that he still wants to return to Facebook social media. Social interaction between people is actually about exchanging and communicating with each other online. If everyone has their own character in the virtual universe, like digital humans, when interacting with them, there will be a very large scene connection. I think this is one of his earliest motivations for creating the metaverse. AI is actually a technology, and the metaverse is more of a scenario. AI is an indispensable technology in the metaverse. AI and the metaverse are not a left or right problem.
Left Hand AI Right Hand Metaverse
At the end of October 2021, Zuckerberg announced that Facebook was officially renamed Meta, which means the metaverse. Zuckerberg also saw this as the next chapter in the company's development. But in 2022, Meta's revenue decreased by 1% year-on-year, and its net profit plummeted by 41%. As for Zuckerberg's decision to expand his metaverse business, Reality Labs, to $13.72 billion in 2022 after losing $10.19 billion in 2021. From a specific business perspective, the sales of its VR hardware Oculus Quest series have decreased, and the metaverse platform Horizon Worlds has not achieved significant growth. In the secondary market, Meta's performance this year has also been quite poor, with its stock price dropping below $90 at one point.
Kwai insiders told reporters: Meta itself is a first tier company, and last year it threw out many big models in one go, from text to visual and audio, including open-source LLaMA, indicating that this field has been making efforts. However, the crux is that Meta has been investing a lot of resources in exploring the metaverse before, but apart from the decent shipment of devices, it has not actually formed a complete application ecosystem. Even after Apple's debut, streaming giant Netflix does not plan to develop corresponding applications The application of the metaverse has not yet formed a consensus on the entire industry at this stage. In addition, the capital market will put some pressure on Meta. In the current situation where traditional businesses such as advertising have a certain impact, it is important and practical to tell the story of AI well
"It can be understood that the metaverse is its long-term or medium to long-term strategic layout, and AI, especially the large model product brought by AIGC, can output profitable market products in the medium and short term. The combination of AI and the metaverse is the future combination. My understanding should be like this. This road should be taken, not only by looking at poetry and distant places, but also by looking at the road under our feet." Zhang Yi said.
Regarding the question of whether Zuckerberg is optimistic about doing metaverse and AI, Zhang Fa'en believes that: The metaverse is a very big concept, although everyone has different understandings, it is definitely becoming more and more digitized and virtualized. Humans will live in a digital world more and more, and will our entire consciousness and thoughts move to this digital world in the future? For example, will the information of our brain neurons become some future collected information, which is actually possible. Therefore, overall, human life is More and more digitized. Zuckerberg's essence in creating the metaverse is to make the entire social world more and more virtual and digitized, and even the entire human world more and more virtual. I think this is a direction, and I also have a positive outlook. I think Zuckerberg is very passionate about his investment in universal AI infrastructure and basic models. In the entire big model industry, Llama is one of the most influential open-source developers
When it comes to his views on the future of AI, Zhang Fa'en believes that AI can actually have a huge impact in all fields. What we have already felt, whether it is writing, designing, or even programmers writing code, let alone talk about intelligence, biomedicine, and education in our current industrial manufacturing industry. AI will bring a transformative effect everywhere.