Meta Platforms Inc. has reportedly recruited four former OpenAI researchers to its newly launched superintelligence lab.

TechCrunch today cited a source as saying that the hires include Trapit Bansal, who joined the ChatGPT developer in 2022. He reportedly played a key role in launching OpenAI’s reinforcement learning program. Reinforcement learning is an AI training method that lends itself to building reasoning models.

The three other OpenAI researchers who have reportedly joined Meta are Lucas Beyer, Alexander Kolesnikov and Xiaohua Zhai. The Wall Street Journal cited sources as saying that the trio helped establish the company’s office in Zurich late last year. They earlier worked at Alphabet Inc.’s Google DeepMind machine learning research lab.

The hires come a few weeks after word first emerged that Meta is forming a superintelligence research group. The lab will be tasked with developing AI models that can outperform humans across a wide range of tasks. The company reportedly launched the unit in response to issues with Llama 4 Behemoth, an internally developed large language model that it previewed earlier this year. Meta has had to delay the algorithm’s launch over concerns about its performance.

Last week, OpenAI revealed that the Facebook parent had tried to poach some of its employees with up to $100 million sign-up bonuses. According to Journal, several of the ChatGPT developer’s researchers have declined Meta’s offers. In some cases, OpenAI reportedly offered staffers “more money and scope to stay.”

Meta is not limiting its recruiting push to OpenAI’s workforce. Earlier this month, the company hired Alexandr Wang, chief executive of AI training dataset provider ScaleAI Inc. Meta also invested $14.3 billion in the startup for a 49% stake.

More high-profile hires could be announced in the coming weeks.

Bloomberg reported today that Meta is in advanced talks to acquire PlayAI Inc., a Palo Alto, California-based voice AI developer. It’s backed by about $21 million in funding from Y Combinator and other institutional investors. The deal is expected to see the Facebook parent hire several of PlayAI’s employees.

The startup provides a cloud platform for creating voice AI agents. The software is powered by two custom AI models known as Play AI 3.0 mini and Dialog. The former algorithm is optimized for hardware-efficiency, while the latter features ten times as many parameters and offers higher output quality.

Meta is reportedly also seeking to hire tech investor Daniel Gross and former GitHub CEO Nat Friedman. The duo currently works at Safe Superintelligence Inc., a startup they co-founded last year with former OpenAI chief scientist Ilya Sutskever. Like Meta’s new superintelligence unit, the company is seeking to develop AI models that can perform many tasks better than humans.

The Facebook parent will spend up to $65 billion this year on data center infrastructure to support its AI programs. As part of the initiative, Meta reportedly plans to build a sprawling new data center with more than 1.3 million Nvidia graphics cards. 

Photo: Wikimedia Commons