Meta's $14.3B deal with Scale AI showing signs of strain
What's the story
Meta's $14.3 billion investment in data vendor Scale AI and the subsequent appointment of CEO Alexandr Wang to lead Meta Superintelligence Labs (MSL) is already showing cracks. Ruben Mayer, Scale AI's former Senior Vice President of GenAI Product and Operations, who was brought over by Wang to help run MSL, has left Meta after just two months, according to TechCrunch.
Leadership changes
Mayer wasn't selected for TBD Labs
During his short stint at Meta, Mayer managed artificial intelligence (AI) data operations teams and reported to Wang. However, he wasn't selected for TBD Labs, the main unit focused on creating AI superintelligence. Meanwhile, TBD Labs is said to be using third-party data vendors other than Scale AI for training its upcoming AI models. These include Mercor and Surge, two of Scale AI's biggest competitors.
Quality issues
TBD Labs researchers have shown preference for Mercor
Despite Meta's multi-billion-dollar investment, sources have claimed that researchers in TBD Labs consider Scale AI's data to be of low quality. They have also shown a preference for working with Surge and Mercor. This is surprising as AI labs usually work with multiple data vendors but rarely invest heavily in one. Scale AI's initial business model relied on a large, low-cost workforce for simple data annotation tasks.
Market shift
Need for high-quality data has increased
As AI models have become more complex, they now need highly-skilled domain experts to create and curate the high-quality data necessary for improving their performance. Scale AI has tried to attract these experts with its Outlier platform. However, competitors like Surge AI and Mercor are growing rapidly as their business models were built on a foundation of high-paid talent from the start.
Strategic moves
Meta is using competitors' data despite investing in Scale AI
A Meta spokesperson has denied any quality issues with Scale AI's product, TechCrunch reports. However, the company is working with third-party data vendors to train its upcoming AI models. This suggests that despite its massive investment in Scale AI, it isn't putting all its eggs in one basket. On the other hand, shortly after losing OpenAI and Google's business, Scale AI laid off 200 employees from its data labeling business due to "shifts in market demand."
Future prospects
New talent expresses frustration navigating the bureaucracy of big company
Scale AI's new CEO, Jason Droege, said the company would hire in other areas of the business after laying off employees. Meanwhile, Meta's AI unit has become increasingly chaotic since bringing on Wang and top researchers from OpenAI and Scale AI. New talent has expressed frustration with navigating the bureaucracy of a big company, while Meta's previous GenAI team has seen its scope limited.
Talent turnover
MSL on track to launch next-gen model by year end
Several members of Meta's GenAI unit have left due to the changes, including MSL AI researcher Rishabh Agarwal. He announced his departure on social media, saying he was compelled by the pitch from Mark Zuckerberg and Wang to build in the Superintelligence team. Despite these departures, MSL has already started working on its next-generation AI model with a planned launch by year's end.