A philosophical divergence between Meta CEO Mark Zuckerberg and Chief AI Scientist Yann LeCun regarding artificial intelligence strategy and timelines became evident last week with the announcement of Meta Superintelligence Labs, generating uncertainty about the company’s future AI direction.

This division within Meta’s AI teams centers on fundamental approaches to AI development. Yann LeCun, Meta’s chief AI scientist and a Turing Award recipient, has publicly asserted that achieving even “cat-level intelligence” remains “very far” from current capabilities. His perspective stands in contrast to Zuckerberg’s recent commitment of $14.3 billion towards the pursuit of superintelligence.

The contradiction became pronounced last week when Zuckerberg announced the formation of Meta Superintelligence Labs, a new division to be led by Alexandr Wang of Scale AI. Zuckerberg conveyed in a memo that the development of superintelligence was “coming into sight” and characterized it as “the beginning of a new era for humanity.” This declaration positions Zuckerberg at odds with LeCun, who has consistently argued over recent years that existing AI methodologies cannot attain even animal-level intelligence, let alone the advanced AI capabilities Zuckerberg has committed to developing.

The philosophical disagreement extends beyond the timeline for AI development and encompasses Meta’s core AI strategy. LeCun has consistently advocated for open-source AI, maintaining its essential role in fostering diversity and democracy within the field. As recently as last year, he commended Zuckerberg’s dedication to open source, stating on LinkedIn that “AI platforms must be open, just like the software infrastructure of the Internet became open.”

LeCun argued that open source facilitates “more diversity in languages, cultures, value systems, and centers of interest in AI assistants.” However, Zuckerberg’s superintelligence memo notably omits any mention of open source. This omission is significant, particularly given that Zuckerberg had written in July of the previous year that “Open Source AI is the Path Forward.” According to information reported by the New York Times, Meta executives even discussed the possibility of “de-investing” in Llama, their open-source model, and potentially adopting closed models offered by competitors such as OpenAI and Anthropic. LeCun continues to promote the open-source message on social media, having recently retweeted content praising Meta’s “open release of Llama that changed the field,” even as his CEO appears to be re-evaluating this strategic direction.

The establishment of Meta Superintelligence Labs reflects Zuckerberg’s initiative to address what had become apparent as a lagging effort in AI. His dissatisfaction reached a critical point in April when the performance of the latest Llama models fell below expectations. The New York Times reported that Zuckerberg discovered his AI team had employed customized benchmarks designed to make the models appear more advanced, a revelation that reportedly displeased the CEO, who had not been adequately briefed on these testing manipulations. This incident underscored a deeper issue: despite Meta’s decision to open-source Llama to encourage widespread adoption, the Chinese startup DeepSeek developed superior models building upon Llama’s foundational technology, effectively out-innovating Meta using Meta’s own contributions.

Tensions have also surfaced in Meta’s broader AI hiring strategy. Zuckerberg has personally contacted more than 45 researchers from OpenAI, presenting compensation packages reportedly reaching as high as $100 million. Wired reported that Meta extended at least 10 “staggeringly high offers” to OpenAI staff members. One high-ranking researcher was offered the role of chief scientist, which is LeCun’s current title, but declined the offer. This intensive hiring drive further suggests Meta’s commitment to the large language model approach, a methodology that LeCun has openly criticized. While LeCun advocates for “world models” that learn through observation of the physical world rather than solely processing text, the newly established superintelligence lab is predominantly staffed by researchers with experience in developing text-based systems like GPT-4.

LeCun’s responses on X (formerly Twitter) regarding his role provide insight into his efforts to maintain influence while acknowledging new organizational realities. When questioned about his position following Wang’s appointment as Chief AI Officer, LeCun affirmed his continued role as Chief AI Scientist and dismissed suggestions that he should lead the superintelligence effort.

Artificial Superintelligence has always made sense as an aspiration and long-term goal.
It has always been FAIR’s long-term goal (as well as mine). It still is, now more than ever.

Artificial General Intelligence never made sense: it’s meant to designate “human-level AI”, but…

— Yann LeCun (@ylecun) July 1, 2025

He stated, “I don’t like managing things and I’m not good at operational stuff. I’m much better at scientific leadership: I lead with ideas.” However, when pressed on X about Meta’s new focus on superintelligence, LeCun carefully navigated his response, stating that “Artificial Superintelligence has always made sense as an aspiration and long-term goal. It has always been FAIR’s long-term goal (as well as mine). It still is, now more than ever.” FAIR is Meta’s existing AI research arm, which LeCun has directed since joining the company over a decade ago.

The existence of these competing visions creates an uncertain future for Meta’s AI endeavors. The company is currently pursuing both approaches simultaneously, with the expectation that one will achieve a breakthrough while the other maintains scientific credibility. However, Zuckerberg’s establishment of a new superintelligence lab, appearing to diverge from his chief scientist’s research philosophy, indicates that Meta has not yet definitively committed to a single path forward.

Featured image credit