(Bloomberg) — OpenAI is phasing out the work it does with data-labeling startup Scale AI, cutting ties with the company days after Meta Platforms Inc. invested billions of dollars in it and hired its founder.
Most Read from Bloomberg
Scale accounted for a small fraction of OpenAI’s overall data needs, according to an OpenAI spokesperson who confirmed the firm’s decision to phase out work with the company. The ChatGPT maker was already in the process of winding down its reliance on Scale before Meta, an OpenAI competitor, took a 49% stake in the firm, the spokesperson said, adding that OpenAI had been seeking other providers for more specialized data needed to support increasingly advanced artificial intelligence models.
OpenAI’s plans inject new uncertainty into Scale’s business in the wake of Meta’s unusual deal. Meta is investing $14.3 billion in Scale and has poached the startup’s chief executive officer, Alexandr Wang, for a new so-called “superintelligence” unit, focused on building a more powerful, and hypothetical, form of AI software. Other Scale employees are expected to follow Wang to Meta to work on AI.A Scale AI spokesperson declined to comment.
Founded in 2016, Scale signed up prominent customers, including Alphabet Inc.’s Google, Meta and OpenAI, providing them with the data needed to build AI models. However, Meta’s deal with Scale raised concerns that the social-media company may gain new visibility into its rivals’ AI development efforts. Google plans to cut ties with Scale, Reuters reported, citing unnamed people familiar with the matter.Right after the Meta deal was announced, OpenAI Chief Financial Officer Sarah Friar had signaled that the company intended to keep working with Scale. “We don’t want to ice the ecosystem because acquisitions are going to happen,” Friar said at the VivaTech conference in Paris last week.Over the past six to 12 months, however, OpenAI had determined that Scale was not the best fit for it because the AI developer needed more data expertise than Scale could provide, the OpenAI spokesperson said. OpenAI has shifted to building more advanced AI models that can mimic the process of human reasoning, as well as agent-like models that can carry out tasks with limited input from users. Forbes previously reported OpenAI had been winding down its Scale work for months.Scale initially focused on working with an army of contractors to do the grunt work of labeling text and images for earlier AI systems. Scale has gradually enlisted better-paid contractors with doctorates, nursing and other advanced degrees to help develop more sophisticated models.
Source link