Tech Tonic | What is Meta’s play with the Superintelligence Labs?

The goal is audacious, developing AGI that can match and eventually exceed human performance, across wide range of tasks, pushing towards true superintelligence

The 28-year old Alexandr Wang, an MIT dropout who founded artificial intelligence company Scale in 2016, is now leading Meta’s AI efforts. This isn’t just any other typical Silicon Valley hiring, and most certainly the age has nothing to do with why I am writing about his role in Meta’s AI vision. In the past couple of days, since Meta Platforms Inc. Chief Executive Officer Mark Zuckerberg announced a major restructuring (this is more substantial than changing the name from Facebook to Meta), the X feed has looked more like how a football club’s feed would, when the transfer window is open to frenetic activity. Anyone who has achieved anything in the AI domain thus far, seems to be joining Meta. And are posting about it. I would presume, for good money too, but I digress.

Mark Zuckerberg has embarked on a major restructuring at Meta, amidst competition with OpenAI and Google. (Official photo)

Trapit Bansal was previously with OpenAI where he co-created the o-series models. Jiahui Yu has also worked at OpenAI where he was part of the core team that created the o3, o4-mini, GPT-4.1, and GPT-4o models, and led perception and multimodal efforts. Yu must have worked alongside Ji Lin on these models. Joel Pobar re-joins from Anthropic. Jack Rae joins from Google DeepMind, having worked there on pre-training for the Gemini 2.5 model. As does Pei Sun, where he worked at post-training and reasoning for Gemini models, as well as the Waymo models. These are just some high-profile names. If you think Meta is trying to recreate the Los Galacticos era, akin to the Real Madrid football club in the 2000s, you’d not be off the mark.

The question really remains — this latest strategic move in artificial intelligence, which represents one of the most ambitious and expensive bets in the company’s history, borne more out of Zuckerberg’s frustration with Meta’s less than ideal position in an AI race that also features OpenAI, Google, Anthropic and Perplexity? There have been sparks of brilliance along the way, such as the Meta AI integration in the Ray-Ban Meta AI glasses (there are new ones with Oakley coming too, called Performance AI glasses), which Google is looking to counter with the expected arrival of Android XR later this year. One would have expected WhatsApp, Instagram and Facebook’s user base to give Meta AI, based on the company’s Llama models, a sort of springboard in terms of active usage that OpenAI’s ChatGPT, Google’s Gemini and Anthropic’s Claude didn’t have.

Unfortunately, it didn’t really pan out as logic may have dictated. Big name hirings, lots of money being thrown at the problem. Meta is also making a $14.3 billion investment in artificial intelligence company Scale, alongside getting its CEO Alexandr Wang, to lead their reconfigured AI project. Secondly, the focus on the word “Superintelligence” may mean Meta is now looking at artificial general intelligence, or AGI, as core to any AI progress. And not a side-project. Till now, Meta’s approach kept two parallel lines open, but that’s no longer the case.

The goal is audacious — developing AGI that can match and eventually exceed human performance, across a wide range of tasks, which would signify push towards true superintelligence. This objective goes well beyond current AI capabilities, which seem to excel at specific tasks but lack flexible reasoning and general intelligence that characterises human cognition.

Meta’s Superintelligence Labs initiative has several important implications for the broader AI industry. First, it validates the growing consensus that AGI and superintelligence represent the ultimate competitive battleground in artificial intelligence. Second, it demonstrates that achieving these goals will require massive financial commitments and organisational focus that may be beyond the reach of smaller players. It is unlikely Superintelligence, or AGI, will be delivered in a DeepSeek-esque moment. Yet, there is no guarantee the collective approach will succeed.

That’s where we must discuss the challenges. Despite what is clearly an ambitious vision backed by substantial resources, Meta’s superintelligence play must square up to a reality in which technical hurdles towards achieving AGI and the risks that will arise subsequently, still remain poorly understood. Is there any guarantee that big money investment and hiring top talent will translate directly to breakthrough results? It is not as if Meta’s main AI rivals are standing still on the matter of AGI. They too have put in significant investments. Case in point, in March, OpenAI announced a fresh $40 billion funding at a $300 billion post-money valuation, towards building AGI.

That is before we even get to the paradigm of regulatory and safety considerations which will inevitably dictate superintelligence development. We’ve not even gotten started on the conversation, and within the AGI development, some or many aspects may need to be reworked in entirety depending on how regulators structure the guidelines. How quickly any AI company is able to respond to those, and indeed the winds of change that keep changing direction regularly, will define the ultimate winner.

Whether this bold gambit pays off remains to be seen, but Meta’s moves (and those in quick time, mind you) have already begun to alter the competitive dynamics of an AI industry where stakes are already high. It is a race toward superintelligence, where Meta’s willingness to commit this level of resources and talent, may either prove visionary, or history may remember it as a cautionary tale. We know Real Madrid’s Los Galacticos investments were the latter. Success isn’t guaranteed, but Mark Zuckerberg is surely trying.

Vishal Mathur is the Technology Editor at HT. Tech Tonic is a weekly column that looks at the impact of personal technology on the way we live, and vice-versa. The views expressed are personal.