Cracks are forming in Meta’s partnership with Scale AI

It has only invest in Meta since June $ 14.3 billion Alexandra Wang, the CEO on the Data dealer Scale AI, and several top executives of startups have been brought to the Meta Superintelligence Labs (MSL). However, the relationship between the two companies has already shown signs of disappointment.

He brought at least one Wang to help run the MSL – Ruben Mayer, a former senior vice -president of Scale AI’s Jenny Product and Operations, told Techcunch that the company had left the meta with only two months later.

Myer spent about five years with Scale AI across two stints. Within a short period of time in Meta, Mayer supervised the AI ​​Data Operation teams and informed Wang, but was not tap to join the company’s TBD labs – the main unit in charge of the AI ​​superintendent, where the top AI researchers from OpenAI landed.

Myer did not respond to two separate requests to comment on TechCrunch.

Further, TBD labs are familiar with this topic, according to five people, working with third -party data vendors other than scale AI for training. Among these third -party vendors are markers and surgeons, two of the largest competitors of AII, people said.

Although AI labs usually work with a number of data vendors – it has been working with Meta Marks and Surge before spending TBD labs – it is rare for an AI lab to invest in a data vendor. This makes this situation particularly significant: even after Meta’s multi-billion investment, a number of sources say that researchers in TBD labs view scale AI data as low quality and have expressed priority for working with enthusiasm and marker.

Scale AI initially created its business on a crowded model that uses a large, low -cost worker to handle general data taces. However, the AI ​​models have become more sophisticated, they now need high-sighted domain experts-such as physicians, lawyers and scientists to produce and refine high quality data needed to improve their performance.

Although the subject has been removed to attract experts with the external platform of the scale AI, competitors like Serge and Marker are growing rapidly because their business models were built on high-paying talent from the beginning.

A Meta spokesman argued about the issue that the scale AI product has quality problems. Surge and Markor refused to comment. Asking about deepest dependence on Meta on competitive data suppliers, Scale AI’s spokesperson led to TechCrunch Initial announcement Meta investment in Startup, which quotes the expansion of companies’ commercial relations.

Meta’s contract with third -party data vendors is probably not probably not keeping all its egg scale AI, even after billions of investments in the beginning. The same can be said about scale AI. Shortly after the announcement of its huge investment with Meta Scale AI Open And Google They said they would stop working with the data supplier.

Shortly after losing these customers, the scale has exceeded AI Its data labeling is 200 employees In July, the business, the company’s new chief executive officer, blamed the change in the “market demand shift” with Jason Drawage. Drage says scale AI will worker in other parts of the business, including government sales – the company has just landed a landing 99 million dollars contract With the US Army.

Some initially assumed that Mater’s investment on Scale AI was really to greed Wang, who has worked in AI space since its establishment in AI 20 2016 and he seems to help Meta to attract top AI talent.

Except Wang, there is an open question around how valuable the meta is.

An current MSL employee has said that several scale executives brought to Meta are not working on the original TBD Labs squad like Maira. Further, scales do not depend on the scale AI exclusively for the Meta Data labeling work.

Meanwhile, two former employees and one current MSL employee said that since Wang and top researchers brought a wave of the Mater AI unit, the Mater AI unit has become gradually chaotic. The new talent of OpenAI and Scale AI has expressed disappointment by navigating the bureaucracy of a large company, while on the other hand, Meta’s previous Jenny team has limited its opportunity, they say.

Tensions indicate that Meta’s largest AI investment till today could stop at a rocky beginning, yet it was supposed to resolve the company’s AI development challenges. After its unnecessary launch Lama 4 In April, Meta CEO Mark Zuckerberg was disappointed with the company’s AI team, a current and a former employee told TechCrunch.

OpenAI and Google began to campaign to turn things with OpenAI and to hit Zuckerberg deals and hiring AI talent at the top.

Wang, Zuckerberg has been able to pull from the top AI researchers OpenGoogle Dipmind and anthropological. AI Voice Startups including Meta also earned Play AI And WaveForms AIAnd the AI ​​image has announced the partnership with the beginning of the generation, Mid -journeyThe

In order to strengthen its AI ambitions, Meta recently announced a number of huge data centers across the United States to build a large one of the largest one $ 50 Billion Data Center Louisia, known as Hyperion, was named after a Titan in the Greek mythology, which was born of the god of the sun.

Wang, who was not AI researcher by the background, was seen as somewhat Unattractive choice To lead an AI lab. Zuckerberg is reported to have discussed more traditional candidates to lead this effort as the chief research officer of the Opehenai Mark chainAnd attempted to achieve its starting Eliaia And MeraThe They all denied.

There are recently some new AI researchers brought from OpenAI Already left metaThe wire has been previously reported. Meanwhile, many members of the Meta Jenny unit have gone in the light of change.

MSL AI Researcher Ish Shov Agarwal is one of the lasts, Post the X This week he will leave the company.

Agarwal said, “Mark and @Alexander_wang’s pitch was incredibly compulsory for the formation of the Superintelism squad,” Agarwal said. “But I finally chose to follow Mark’s own suggestion: ‘The biggest risk you can take in a world that is changing so fast is not taking any risk'” “

Later, Meta asked about his time and decided to leave his decision, Agarwal refused to comment.

The director of product management for the generator AI, ShadowAnd the research engineer, Rohan confirmedIt has also announced their departure from Meta in recent weeks. Now the question is whether the meta can stabilize its AI operations and maintain the talent needed for future success.

MSL has already started working on its next generation AI model. According to the report from close Business insiderThis is the goal of turning it on the end of this year.