Mark Zuckerberg is working on developing artificial general intelligence in Meta

Mark Zuckerberg set a new goal for himself in the tech field: creating artificial general intelligence (AGI). The CEO of Meta, formerly known as Facebook, has moved his FAIR (Facebook AI Research) research group to work with a team developing generative AI products in its social media applications.

This strategic move seeks to integrate the company’s AI advancements to directly benefit its billions of users.As published the verge, In addition, Zuckerberg announced that by the end of this year, Meta will have more than 340,000 Nvidia H100 GPUs, becoming one of the main players in the accumulation of computing power needed for training and running large-scale AI models. I will stand.

“We came to the conclusion that to build the products we want to build, we have to build general intelligence,” Zuckerberg said in an exclusive interview. the verge, “I think it’s important to point out because many of the best researchers want to work on the most ambitious problems.”

The Target CEO’s commitment to AI reflects not only structural changes within his company, but also the intense competition for talent in this area. Hiring researchers and AI experts has become a war for talent, where experts can command salaries in the millions of dollars. This competition has intensified due to the presence of many companies seeking similar profiles and a large amount of venture capital willing to finance various initiatives in the sector.

Meta’s CEO admits that he does not have a precise definition of artificial general intelligence (AGI) nor does he know when it will arrive. “It can be argued whether general intelligence is similar to human intelligence, or if it is more human, or if it is a superintelligence from the distant future. But to me, the important thing is its breadth, meaning intelligence has all these different abilities where you have to be able to reason and have intuition,” he told the outlet.

One of the most important developments for Zuckerberg is the ability to generate code in large-scale language models (LLM), although it was not initially considered important for meta applications. Although the skill was developed within the Llama 2 project, its structural value for understanding and reasoning about models was later recognised.

“One hypothesis was that coding is not that important because not many people will ask coding questions on WhatsApp,” he said. “It turns out that coding is really important from a structural perspective so that LLMs are able to understand the rigor and hierarchical structure of knowledge and, in general, have a more intuitive understanding of the logic,” he explained.

Meta is currently working on Llama 3 in hopes of incorporating these advanced code generation and logic capabilities, following in the footsteps of models like Google’s Gemini. The ambition of Llama 3 and future models is to reach and lead the state of the art within the AI ​​industry. This development reinforces Meta’s commitment in the race to develop artificial intelligence that matches or even surpasses human intelligence in terms of intellectual capacity.

Amid tensions raised worldwide by the development of AI, the executive reaffirmed his commitment to an open source model for Llama, a technology that is considered open source, except in a few cases. This situation reverses OpenAI’s trend of restricting its initial transparency. Zuckerberg, who has significant control over Meta due to his shareholding influence, has raised concerns about the concentration of technological power and has advocated for more equitable access.

Zuckerberg’s commitment to open AI development is not without criticism. Some observers, such as Sam Altman, propose a more closed approach to security in the development of AI. However, according to Zuckerberg, existing AI models implemented by leading companies in this field have not yet caused irreparable harm. Meta’s CEO has identified calls for greater control over AI as a potential strategic play that matches the interests of the dominant companies in the market, despite perceived concerns over security.

Meta benefits from a huge user base and a highly profitable social media business, elements that can be strengthened by advanced AI capabilities. By effectively opening up its models, Meta can have an even greater impact on the tech ecosystem. Despite this, Zuckerberg has not yet decided whether the final AGI developed by Meta will be open source, believing that any future commitment should be flexible and responsive to circumstances.