Meta Appoints Four Former OpenAI Researchers to Superintelligence Lab

Meta Platforms Inc. has reportedly hired four former OpenAI researchers to join its newly established superintelligence laboratory.

According to TechCrunch's latest report, a source revealed that the hires include Trapit Bansal, who joined ChatGPT's development team in 2022. He played a pivotal role in launching OpenAI's reinforcement learning projects - a key AI training method for building reasoning models.

The other three ex-OpenAI researchers joining Meta are Lucas Beyer, Alexander Kolesnikov and Xiaohua Zhai, as revealed by Wall Street Journal sources. The trio helped establish the company's Zurich office last year and previously worked at Google DeepMind under Alphabet Inc.

These appointments follow early reports about Meta's efforts to form a superintelligence research team. The lab aims to develop AI models capable of outperforming humans across diverse tasks. The department's creation responds to challenges with Llama 4 Behemoth - an internally developed large language model recently previewed by Meta which required delayed release due to performance concerns.

Last week, Openai disclosed Facebook's parent company attempted to lure some employees with $100 million signing bonuses. According to WSJ reports, several ChatGPT developers rejected Meta's offers. In some cases, Openai allegedly provided "more financial support and development space" to retain staff.

Meta's recruitment strategy extends beyond OpenAI. Earlier this month, the company appointed Alexandr Wang, CEO of AI training data provider ScaleAI Inc. The investment includes $1.43 billion for 49% equity in the startup.

More high-profile hires could be announced in upcoming weeks.

Bloomberg reported today that Meta is engaging in advanced negotiations with Palo Alto-based voice AI developer PlayAI Inc. This startup secured $21 million from Y Combinator and other investors. The deal would enable Facebook's parent company to hire several PlayAI employees.

The startup provides a cloud platform for creating voice AI agents powered by two customized models: Play AI 3.0 mini and Dialog. The former algorithm optimizes hardware efficiency while the latter features tenfold parameter count for enhanced output quality.

Meta is also seeking to recruit tech investor Daniel Gross and former GitHub CEO Nat Friedman. Both currently work at Safe Superintelligence Inc., the startup they co-founded with former OpenAI chief scientist Ilya Sutskever in 2023. Similar to Meta's new superintelligence division, this company focuses on developing AI models that surpass human capabilities in multiple domains.

Facebook's parent company plans to invest up to $65 billion in data center infrastructure this year to support AI initiatives. As part of this strategy, Meta reportedly plans to build a massive data center with over 1.3 million Nvidia graphics cards.