Meta invests in the AI ​​company scale and recruits its CEO for the “superintelligence” team

Meta makes an investment of $ 14.3 billion in the scale of artificial intelligence companies and recruiting its CEO Alexandr Wang to join a team developing a “superintendent” to the technology giant.

The agreement announced Thursday reflects a push of the meta-PDG Mark Zuckerberg to revive AI efforts in the Facebook and Instagram parent company, because it faces difficult rivals like Google and Openai.

Meta announced what he called a “strategic partnership and an investment” with Scale on Thursday evening. Scale said that the investment of $ 14.3 billion puts its merchant value to more than $ 29 billion.

Scale has said that it will remain an independent company, but the agreement will considerably expand the Meta scale and trade relationship “. Meta will hold a 49% stake in the startup.

Wang, although on Meta with a small group of employees from other scales, will remain on the board of directors of Scale. The replacement is a new interim CEO Jason Droege, who was previously the director of the company’s strategy and had played management positions at Uber Eats and Axon.

The emphasis is increasingly putting on Zuckerberg on the abstract idea of ​​”superintelligence” – that rival societies call for general artificial intelligence, or AGE – is the last pivot for a technological leader who, in 2021, went all to the idea of ​​metavers, changing the name of the company and investing billions of billions to advance the virtual reality and related technologies.

It will not be the first time, because the beginnings of Chatgpt in 2022 have sparked an arms race for AI that a large technological company has engulfed talents and products in innovative AI startups without officially acquiring them. Microsoft hired the key staff of the IA inflection startup, including the co-founder and CEO Mustafa Suleyman, who now directs the AI ​​division of Microsoft.

Google attracted the leaders of the Chatbot company IA Character.ai, while Amazon concluded an agreement with ADEPT based in San Francisco which sent its CEO and key employees to the electronic commerce. Amazon also obtained a license for AIPT systems and sets.

Wang was a 19-year-old student at Massachusetts Institute of Technology when he and co-founder Lucy Guo began the scale in 2016.

They won an influential support that summer from the startup incubator y combinator, which was led at the time by Sam Altman, now CEO of Openai. Wang abandoned it, following a trajectory similar to that of Zuckerberg, who left Harvard University to start Facebook more than a decade earlier.

Scale’s argument was to provide human work necessary to improve AI systems, hiring workers to draw boxes around a pedestrian or dog in a street photo so that autonomous cars can better predict what is in front of them. General Motors and Toyota were among the customers of Scale.

What a scale offered to AI developers was a more personalized version of Amazon Mechanical Turk, which has long been an essential service for self -employed workers with temporary online jobs.

More recently, the growing marketing of AI -language models – technology behind the Openai Chatppt, Google Gemini and Meta Lama – has brought a new market for SCALE annotation teams. The company claims to serve “all the main models of great language”, including Anthropic, Openai, Meta and Microsoft, helping to refine their training data and test their performance. We do not know what the meta-affair will mean for the other SCALE customers.

Wang has also sought to establish close relations with the US government, winning military contracts to provide AI tools to the Pentagon and attend the inauguration of President Donald Trump. Trump’s science and technology office manager Michael Kratsios was a large -scale framework for the four years between Trump’s first and second mandates. Meta also started providing AI services to the federal government.

Meta has adopted a different approach to AI than many of its competitors, freeing its flagship lama system for free as an open-source product that allows people to use and modify some of its key components. Meta says that more than a billion people use its AI products each month, but it is also largely considered to be lagging behind compared to competitors such as Openai and Google to encourage the use of consumers of large languages ​​models, also called LLMS.

He has not yet published his allegedly the most advanced model, Llama 4 Behemoth, despite the preview in April as “one of the most intelligent LLMs in the world and our most powerful to date.”

The chief scientist of Meta AI, Yann Lecun, who, in 2019, was a winner of the first Prize of Computer Science for his Pioneer Work of AI, expressed his skepticism as for the current accent of the technological industry on models of great language.

“How do we build AI systems that understand the physical world, which have a persistent memory, which can reason and plan?” Lecun asked at a conference on French technology last year.

These are all characteristics of intelligent behavior that large languages ​​”can not do, or they can only do them in a very superficial and approximate way,” said Lecun.

Instead, he stressed Meta’s interest in “following a path to AI systems on a human level, or perhaps even superhuman”. When he returned to the Vivatech annual conference of France on Wednesday, Lenun dodged a question on the pending scale agreement, but said that the plan of his research team on AI had “always been to reach human intelligence and go beyond”.

“It’s just that now we have a clearer vision in how to do this,” he said.

Lecun co -founded the AI ​​research division of Meta over ten years ago with Rob Fergus, another professor at New York University. Fergus left later for Google but returned to Meta last month after an absence of 5 years to manage the research laboratory, replacing the longtime director Joelle Pineau.

Fergus wrote on LinkedIn last month that Meta’s commitment to long -term research of the AI ​​”remains unshakable” and described work as “the creation of human level experiences which transform our way of interacting with technology”.