AI: Meta's 'SuperIntelligence' team gets settled. RTZ #817
Three months after Meta’s flurry of Global AI Talent hiring spree with Scale AI and from other LLM AI companies like OpenAI, Apple, Google and others, the form of its new AI ‘Superintelligence’ super team is becoming clear. All chomping at the bit to deliver Meta AI powered AI products and services to its 3.4+ billion daily active users, and millions of advertisers.
Catalyzed at the highest levels by founder/CEO Mark Zuckerberg, the multi-billion dollar AI Dream Team seems to be organized to focused on AI Research (two labs), AI Applications and AI Infrastructure opportunities.
The Information gives us the latest upate on the Superintelligence Organization in “Meta Plans Fourth Restructuring of AI Efforts in Six Months”:
“Meta Platforms is planning the fourth restructuring of its artificial intelligence efforts in six months, according to three people familiar with the matter.”
If it’s the fourth restructuring, it appropriately comes in four parts: two AI Research labs, one AI products focused group, and a group working on AI Infrastructure:
“Meta’s new AI organization, Meta Superintelligence Labs, is expected to be divided into four groups: a new lab that has been going by TBD Lab, short for to be determined; a team focused on products that include the Meta AI assistant; a team focused on infrastructure; and the company’s Fundamental AI Research lab, which works on longer-term research, two of the people said.”
The products focus seems to echo what OpenAI is also doing with its new ‘CEO of Applications’, to build consumer facing AI applications and services in the AI Tech Stack Box 6 below.
“The changes are coming after stumbles earlier this year set off a particularly turbulent period for Meta and its employees working on AI. Meta CEO Mark Zuckerberg has gone to great lengths to overhaul that work. Earlier this summer, his company agreed to spend billions of dollars to hire former Scale AI CEO Alexandr Wang and former GitHub CEO Nat Friedman to co-lead Meta Superintelligence Labs. Meta has also been on a recruiting blitz, hiring dozens of researchers from rivals including Google, OpenAI and Anthropic.”
These plans are of course preliminary and fluid, highlighted by the ‘TBD’ nature of the first unit.
“Meta’s plans for the restructuring haven’t been announced internally and could change.”
“TBD Lab, which is working on the newest version of Meta’s flagship large language model Llama, is expected to have several leads, the two people said.”
A big question is whether Llama will continue to be open source, or potentially go closed like OpenAI, Anthropic and others. This is after Meta lost some of its global open source mojo to DeepSeek and other AI companies in China. OpenAI also recently announced its open weights initiatives.
“Jack Rae, formerly of Google, is expected to lead pre-training, according to two of the people familiar with the matter. During pre-training, a model learns how to predict text from trillions of words gathered from the internet and other datasets.”
Meta’s well-publicized hires from Apple’s senior AI ranks is focused on AI infrastructure at the TBD Lab, a notable appointment.
“Ruoming Pang, who previously led model development at Apple, is expected to lead infrastructure for TBD Lab, the two people said. That subgroup is distinct from the broader infrastructure group. Jiahui Yu, formerly of OpenAI, is expected to lead multimedia efforts, which would enable a model to understand and generate content other than text.”
Note the long list of AI researchers with Chinese backgrounds, a broader dynamic of the AI Talent War globally I’ve discussed.
“Among the leaders of post-training is expected to be Hongyu Ren, formerly of OpenAI, and Pei Sun, formerly of Google, according to one of the people familiar with the matter. Post-training involves using more refined training data to help a model improve at specific tasks.”
Nat Friedman is focused on the AI products portfolio around Meta AI, also a notable appointment given his VC background. Again, echoing OpenAI efforts here as well.
“Friedman is expected to continue overseeing products for Meta Superintelligence Labs, the person said. Robert Fergus, who co-founded FAIR, is expected to continue leading that lab. Aparna Ramani, a long-time vice president of engineering for infrastructure, is expected to join the organization as the head of the broader infrastructure group.”
And Meta is putting a long-time Meta engineer, Aparna Ramani in charge of the AI infrastructure side, an area I’ve also discussed at length.
“Meta most recently restructured its efforts in AI in June, shortly after it hired Wang and Friedman. In May, the company cleaved its generative AI group—previously led by Al-Dahle—into teams overseeing AI research and another working on AI products. In February, Meta moved two engineering leaders from the group and installed Loredana Crisan, then head of Messenger, to lead product, in an early sign of the shakeup to come.”
All in all, some useful clarity on Meta’s AI plans, rapidly leveraging its mammoth AI capex investment in data centers and other infrastructure.
This provides a little clarity vs other LLM AI peers at this early point in the AI Tech Wave. Lots of uncertainties ahead at Scale, along with the possible opportunities. Stay tuned.
(NOTE: The discussions here are for information purposes only, and not meant as investment advice at any time. Thanks for joining us here)
