OpenAI is scaling back its contracts with Scale AI, a data labeling startup recently acquired by social media giant Meta.
According to a Thursday report from Bloomberg, the ChatGPT maker is winding down its reliance on the startup’s data, a few days after Meta announced a $14.8 billion deal for a 49% ownership stake in Scale AI.
The acquisition was Meta’s second-largest and will also have Scale CEO Alexandr Wang joining Meta’s experimental AI project. The companies announced the deal on June 12.
Scale AI supplies labeled data used to train and improve artificial intelligence (AI) models. The startup, founded in 2016 and backed by over 100 investors, is a supplier to popular AI companies including Anthropic, Cohere and Adept. According to PitchBook, it raised $100 million in a Series C round in 2019.
Another company reportedly phasing out contracts with Scale is Google. According to Reuters, the move is backed by concerns that the deal could grant Meta insights into competitors’ AI developments.
Related: Elon Musk’s AI firm xAI faces lawsuit over data center air pollution
Related: Amazon to invest $13B in Australian AI data center infrastructure
An OpenAI spokesperson told Bloomberg the company started phasing out its contracts with Scale over the past year, seeking more specialized data for its AI models. The startup accounted for a small portion of OpenAI’s data needs.
In a statement about the deal, Scale interim CEO Jason Droege said that the startup “remains, unequivocally, an independent company,” adding that “nothing has changed about our commitment to protecting customer data.”
OpenAI is now betting on alternative data suppliers for its operations, including newcomers in the sector such as Mercor.
According to Bloomberg, Scale initially hired “an army of contractors” to label images and text for early AI systems over the past few years. Later, the company moved to hire more educated contractors to aid in the development of sophisticated AI models.
Magazine: AI Eye: AI is good for employment says PWC — Ignore the AI doomers
Read the full article here
