Can Scale AI and Alexandr Wang reignite Meta’s AI efforts?

Meta is reported to be investing almost $ 15 billion Data Labeling Farm Scale takes 49% of the AI ​​and Startup, as well as the CEO Alexandra Wang to help lead a new “superjections” lab in the company.

The deal reminds of Meta’s previous large and risky bats Acquisition of $ 19 billion in WhatsApp And of the Purchase of $ 1 billion InstagramThe When these attachments were closed, many people suggested that the meta paid fairly for platforms – and today’s speech is not different. Investors and founders have no shortage of people who were shaking their heads because of Meta’s latest tie-up this weekend.

In the end, WhatsApp and Instagram CEO Mark Zuckerberg has become an integral part of the empire. The question is whether the scale AI agreement will work in the same way, whether Zuckerberg’s presidential technique again proves the strategy, or whether the company has been on the straw in a misleading attempt to catch up to rivals like Open, Google and Anthropics.

In this case, the meta is not betting on any emerging social media application, but top AI models are used in data used for training. Over the past several years, top AI labs like Openai have relied on scale AI to produce and label data used for training models. In recent months, the scale AI and its data vaccine competitors have begun to recruit high quality data for Frontier AI labs, such as PhD scientists and senior software engineers.

It can benefit the meta to have a close relationship with a scale of data as a scale. According to a person known about the matter, Meta leaders complained about the lack of innovation about data in the top AI teams of the company.

Early this year A family of AI model has launched Lama 4 It matches the capacity of the Chinese AI Lab DEPSEC model and it was widely seen as frustration. Without helping the subjects, Meta is trying to deal with an attraction problem. According to information compiled by the signalfire, Meta lost its top talent in 4.3% AI labs In 2024.

Meta is not just betting on the scale AI to re -sort her AI effort, but also in Wang to lead the new superintendent team mentioned. The 25-year-old CEO has proved himself as a powerful startup founder-he is as ambitious as Silicon Valley, a good salesman and very well-known. Wang has been the last few months Meeting with world leaders Discussing the impact of AI in society.

Wang has never led such an AI lab before, however, the AI ​​research background does not do as other AI lab leaders like Elias Sutsvaver or Myster’s Arthur Mansash. This is why the appointment of meta high-profile talent is also said to be Dipmind To make the new AI research team out.

The next fate of the scale AI is somewhat flexible. The role of real-world data in AI model training is changing-some AI labs have come up with efforts to collect data at home, while others have increased their dependence on synthetic (ie AI-imposed) data. In April, the information said Scale AI’s Missing some of the earning goalsThe

According to Robert Nishihar, co-founder of any scale, several Frontier AI labs are exploring the fancy ways of gaining data and exclusion, many of which are quite calculated.

“Data is an ongoing goal,” Nishihar told TechCrunch in an interview. “It’s not just a limited effort to catch it – you have to invent.”

It is possible that Meta and Wang’s relationship can intimidate other AI labs that have worked in the Tradition with the scale AI. If that is the case, this agreement may be a compassion for non -conventional data suppliers such as Turing, Surge AI and even recently launched LM Arena.

Turing CEO Jonathan Siddharth told TechCrunch through the email that he had raised interest from customers in the light of the rumors around Meta’s contract with AII.

“I think there will be some clients who would like to work with a partner that is more neutral,” he said.

Only time will say how Mater’s investment will come out for his AI effort, but the company has a significant basis. Meanwhile, the competition is not decreasing. OpenAI is a model that is ready to publish its first public -available model in its next flagship model, GPT -5, as well as years after years -a model that will compete with Meta’s current and future Lama release.