Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
Meta hired a very influential operator, Trapit Bansal, to work on his reasoning models under the The new unit of superintendent of the company of the companyA person familiar with the question tells of Techcrunch.
Openai spokesperson Kayla Wood confirmed to Techcrunch that Bansal had left Openai. Bansal Linkedin page said he left Openai in June.
Bansal has been working at Openai since 2022 and has been a key player to launch the business work on strengthening learning alongside the co -founder Ilya Sutskever. It is listed as a fundamental contributor on the first model of reasoning of Openai, O1.
Bansal could offer a significant elevator to the AI of Meta superintelligence laboratory, which also presents leaders such as the old Scales that you have CEO Alexandr Wang and is seek to add The former CEO of Github, Nat Friedman, and the co-founder of the Surin-Allocation Safe Daniel Gross. Bansal could help Meta develop a reasoning model on the border AI that is competitive with advanced technology, such as the O3 or the Deepseek R1. Currently, Meta does not offer a model of reasoning has publicly.
In recent months, Mark Zuckerberg has been being hired to build the new Meta AI team, Offer remuneration packages of $ 100 million To the best researchers who join his business. We do not know what Bansal was proposed to join this agreement.
Nevertheless, it seems that Zuckerberg has managed to grasp the research talents on AI.
Three other former Openai – Lucas Beyer, Alexander Kolesnikov and Xiaohua Zhai – have also joined the Meta AI AIA in recent weeks, also joined the Meta IA team, The Wall Street Journal reported Wednesday. Bansal will join them, alongside the former researcher from Google Deepmind, Jack Rae, and the former automatic learning chief at Startup Sesame, Johan Schalkwyk, according to Bloomberg.
To further fill his new AI unit, Zuckerberg would have tried to Acquire startups with heavy AI research laboratoriesLike the safe superintendent of Sutskever, Mira Murati Labs reflection machines and perplexity. However, these talks have never progressed to a last step.
On a recent podcast, the CEO of Openai, Sam Altman, said that Meta had tried to rob the best talents of her startup, but said that “None of our best people decided to take it on this. »»
A Meta spokesperson refused to comment.
The AI reasoning models have a key area for the Meta AI superintelligence team to do well. During the last year, Openai, Google and Deepseek have sent highly efficient AI reasoning models that have pushed the limits of what software can do. By forming AI models to solve problems before giving an answer, using additional time and IT resources to do so, AI laboratories have managed to improve AI performance on real world references and tasks.
The Meta AI superintelligence laboratory could become a key internal group that feeds products throughout the company, much like the Deepmind unit of Google. Meta has an ambitious effort for Build AI agents for business Under the former CEO of SALESFORCE DE L’IA, Clara Shih. In order to build competitive agents, META must develop models of reasoning on border AIs to feed them.
With the addition of Bansal and other key researchers from AI, Meta hopes to advance in the AI race. This can be difficult since OPENAI plans to publish an open AI reasoning model In the coming weeks – an offer that could add even more pressure on the Open IA offers from Meta.
This story has been updated after publication with more details.
