Meta’s AI Superintelligence Supergroup Begins to Take Shape
It’s been a frenzied year in the AI sector and Meta has often found itself right at the centre of the whirlwind.
For several months now, Mark Zuckerberg has been putting together a team of some of the most talented AI engineers and computer scientists in the world, many of whom he has recruited directly from Meta’s competitors and boundary-pushing start-ups.
His vision is for this team to develop the most advanced AI models known to man, which he has described variously as ‘artificial general intelligence’ or ‘superintelligence’.
Now, with the recruitment process seemingly over, Meta has begun structuring its new hires into teams.
This reorganisation, announced internally on 19 August, sees the creation of four distinct groups, each of which will be tasked with separate projects.
"Superintelligence is coming, and in order to take it seriously, we need to organise around the key areas that will be critical to reach it — research, product and infra," wrote Alexandr Wang, Meta's new Chief AI Officer, in an internal memo.
A four-pronged structure
The newly restructured Meta Superintelligence Labs will now comprise four separate divisions, each with distinct responsibilities.
TBD Lab, led by Wang himself, will oversee the development of Meta's LLMs, including the Llama tools that power the company's AI assistant.
The Fundamental AI Research lab, known as FAIR, will continue its decade-long focus on longer-term research projects under the leadership of Rob Fergus, who co-founded the division all the way back in 2014.
A Products and Applied Research team, headed by former GitHub CEO Nat Friedman, will translate research into consumer-facing products.
The fourth division, MSL Infra, will handle the expensive infrastructure requirements needed to support Meta's AI ambitions, with Aparna Ramani leading the team.
His purview will encompass data centres and the physical backbone of Meta’s digital plans.
Behind the billion-dollar recruitment scheme
The restructuring comes after a period of flux for Meta, which has seen several high-profile departures and expensive new hires in the AI department.
Joelle Pineau, a leading Computer Scientist at Meta, left the company earlier this year to join AI start-up Cohere.
Angela Fan, a Research Scientist who worked on Meta's open-source Llama model, recently departed for OpenAI.
Loredana Crisan, a VP of Gen AI, is leaving to join software company Figma as Chief Design Officer.
Meanwhile, Meta has been aggressively poaching talent from its competitors, offering them extraordinary compensation packages to come aboard.
In July, the company appointed Shengjia Zhao, an OpenAI researcher and co-creator of ChatGPT, as its Chief AI Scientist.
Meta has reportedly offered some researchers pay packages reaching hundreds of millions of dollars, with at least two cases involving US$1bn packages over several years.
The company has poached at least 18 researchers from OpenAI and numerous others from Google, according to industry sources.
A shift in strategy
The latest reorganisation reflects deeper strategic changes within Meta's AI approach.
Alexandr Wang's new team has chosen to abandon Meta's previous frontier model, called Behemoth, and start from scratch on a new model after disappointing performance tests delayed its release last spring.
The new team has also discussed making Meta's next AI model ‘closed’, which would represent a major departure from the company's long-time philosophy of open-sourcing its models.
This shift has created internal tensions, with some members of Meta's established AI teams reportedly uncomfortable with the influx of new hires and changing direction.
"Already in the past month, I've seen meaningful progress in each of these collaborations," Wang wrote in his memo regarding the new structure.
Zuckerberg is putting his money where his mouth is
Zuckerberg's determination to compete in the AI race is evident in Meta's massive financial commitments.
The company announced capital expenditures could reach US$72bn this year, with most funds directed towards building data centres and hiring AI researchers.
Meta also invested US$14.3bn in start-up Scale AI as part of bringing Wang aboard as Chief AI Officer.
In an investor call last month, Zuckerberg said he was betting on superintelligence to usher in "a new era of individual empowerment", while noting that AI has already improved Meta's core advertising business.
The restructuring is expected to be the final organisational change for some time, as Meta seeks to stabilise its AI efforts and accelerate development of superintelligence technology to compete with rivals like OpenAI and Google.
Whether these latest changes will enable Meta to achieve its ambitious AI goals remains to be seen, as the company continues navigating the complex challenge of transforming substantial financial investment into technological breakthroughs.
