Meta Superintelligence Team (MSL) is a dedicated research and development division within Meta Platforms, established with the explicit goal of advancing artificial intelligence to achieve superintelligence, a form of AI with capabilities surpassing human intellect. The formation of this team reflects Meta's increased focus on advanced AI development and its role as a competitor in the AI sector. ^[1]

Overview

Meta's push into superintelligence development is an initiative led by CEO Mark Zuckerberg, who has publicly stated his view that superintelligence could mark "the beginning of a new era for humanity." The company is investing significant resources, including billions of dollars in talent acquisition and hundreds of billions in computational infrastructure, to accelerate its progress in this field. This expansion is part of a broader effort to compete with other leading AI research organizations such as OpenAI, Google DeepMind, and Anthropic.

The establishment of the Meta Superintelligence Team reflects a shift in Meta's AI strategy, moving towards more ambitious, long-term research goals. This initiative is designed to consolidate AI talent and resources under a unified vision, with the aim of advancing AI capabilities in areas such as reasoning, problem-solving, and general intelligence. The company's financial backing and strategic partnerships indicate the priority of this endeavor within Meta's overall business objectives. ^{[1] [2] [5] [7]}

Formation and Leadership

The Meta Superintelligence Team (MSL) was formally announced by Mark Zuckerberg on Threads, with details further elaborated in an internal memo. Zuckerberg stated that the overall organization is called Meta Superintelligence Labs (MSL), encompassing all of Meta's foundations, product, and FAIR (Fundamental AI Research) teams, along with a new lab focused on developing the next generation of models. Alexandr Wang, the former CEO of Scale AI, was appointed as Meta's new Chief AI Officer and leads the MSL team. Wang co-leads the new lab with Nat Friedman, the former head of GitHub, focusing on AI products and applied research.

Shengjia Zhao, an engineer formerly with OpenAI, was named the chief scientist for Meta's new Superintelligence Lab. In this role, Zhao sets the research agenda and scientific direction for the lab, working directly with Mark Zuckerberg and Alexandr Wang. ^{[1] [2] [3]}

Talent Acquisition Strategy

Meta has undertaken a talent acquisition campaign to staff its Superintelligence Team, involving significant investment to attract AI researchers and engineers from rival companies. This strategy includes offering competitive compensation packages, with some signing bonuses reported to be as high as $100 million for experienced talent. Total compensation packages for some individuals have reportedly reached hundreds of millions of dollars.

The company has actively recruited individuals from prominent AI organizations, including OpenAI, Google DeepMind, Anthropic, Apple, and Ilya Sutskever's Safe Superintelligence. This recruitment effort has influenced the AI talent market, contributing to increased competition for professionals in the field. ^{[1] [2] [3] [6]}

Key Team Members

The Meta Superintelligence Team includes researchers, engineers, and technical professionals with prior experience at major organizations involved in artificial intelligence research and development.

Shengjia Zhao, formerly involved in the development of ChatGPT at OpenAI, was appointed Chief Scientist of Meta's Superintelligence Lab. Other individuals from OpenAI on the team include Trapit Bansal, Ji Lin, Hongyu Ren, Jiahui Yu, Suchao Bi, Huiwen Chang, Lucas Beyer, Alexander Kolesnikov, and Xiaohua Zhai. Their previous work encompassed areas such as reinforcement learning, voice interfaces, image processing, and large language model development. Trapit Bansal pioneered reinforcement learning on chain of thought and was a cocreator of o-series models at OpenAI. Shuchao Bi cocreated GPT-4o voice mode and o4-mini, and previously led multimodal post-training at OpenAI. Huiwen Chang cocreated GPT-4o's image generation and previously invented MaskIT and Muse text-to-image architectures at Google Research. Ji Lin contributed to o3/o4-mini, GPT-4o, GPT-4.1, GPT-4.5, 40-imagegen, and the Operator reasoning stack. Hongyu Ren cocreated GPT-4o, 4o-mini, o1-mini, o3-mini, o3 and o4-mini, and previously led a group for post-training at OpenAI. Jiahui Yu cocreated o3, o4-mini, GPT-4.1 and GPT-4o, and previously led the perception team at OpenAI and co-led multimodal at Gemini. Shengjia Zhao cocreated ChatGPT, GPT-4, all mini models, 4.1 and o3, and previously led synthetic data at OpenAI. Lucas Beyer, Alexander Kolesnikov, and Xiaohua Zhai, who previously helped launch OpenAI’s Zurich office, were noted to have joined Meta, though their inclusion on specific internal lists was subject to "technicalities."

Former Google DeepMind contributors Jack Rae and Pei Sun are part of the team, along with engineers Tianhe Yu, Cosmo Du, and Weiyue Wang, who were involved with the Gemini model. Jack Rae was a pre-training tech lead for Gemini and reasoning for Gemini 2.5, and led Gopher and Chinchilla early LLM efforts at DeepMind. Pei Sun contributed to post-training, coding, and reasoning for Gemini at Google Deepmind, and previously created the last two generations of Waymo's perception models. From Apple, the group includes Ruoming Pang, who directed work related to large language models, as well as Mark Lee and Tom Gunter.

The team also includes individuals from other AI-focused companies. Joel Pobar was previously affiliated with Anthropic, and had prior experience at Meta for 11 years on HHVM, Hack, Flow, Redex, performance tooling, and machine learning. Johan Schalkwyk worked at Sesame AI, and was a former Google Fellow and early contributor to Sesame, and technical lead for Maya. Alexandr Wang is the founder of Scale AI. Daniel Gross, co-founder of Safe Superintelligence, and Nat Friedman, former CEO of GitHub, are also identified as contributors to Meta's AI initiatives.

These individuals are part of a broader group of 44 professionals reported to be involved in Meta’s Superintelligence efforts. ^{[1] [2] [3] [4] [8]}

Investment and Resources

Meta is committing substantial financial and computational resources to its superintelligence initiative. The company invested $14.3 billion to acquire a 49% stake in Scale AI, a company specializing in labeling data for AI systems, which is considered crucial for training advanced AI models.

Beyond talent acquisition and strategic investments, Mark Zuckerberg has stated that Meta plans to invest "hundreds of billions of dollars into compute" for superintelligence development, leveraging the company's significant capital to fund this undertaking. This investment in computing power is considered necessary for training the large-scale models required for superintelligence development. ^{[1] [2] [6]}

Industry Reaction and Competition

Meta's intensive pursuit of superintelligence and its talent acquisition strategy have generated attention and reactions from across the AI industry. Competitors, particularly OpenAI, have expressed concerns regarding Meta's recruitment efforts. Sam Altman, CEO of OpenAI, described Meta's actions as "somewhat distasteful" and stated that OpenAI is reevaluating its compensation structures to retain its researchers. Altman also reportedly communicated internally that while Meta had acquired "a few great people," they "didn’t get their top people and had to go quite far down their list."

The significant competition for AI talent and resources highlights the high stakes in the field of advanced AI development. Meta's substantial investments and strategic hires position it as a significant participant alongside established leaders like Google DeepMind, OpenAI, and Anthropic, increasing competition in the artificial intelligence sector. ^{[1] [4] [2]}