Why Meta’s AI Pioneer Yang Li‑kun Is Being Marginalized: Power Struggles Behind the Scenes
The article examines how Meta’s CEO Mark Zuckerberg’s aggressive talent‑buying and commercial focus have sidelined Turing‑award winner Yang Li‑kun, detailing the restructuring of Meta’s AI labs, the clash over research directions, and the broader dilemma of balancing academic innovation with business imperatives in the AI industry.
Yang Li‑kun’s Sudden Demotion
One night, Yang Li‑kun, a Turing‑award winning AI pioneer, found himself reporting to a 28‑year‑old newcomer after Meta’s CEO Mark Zuckerberg invested $148 billion to acquire a new AI leader.
Yang, who has served Meta for 12 years, now faces a younger boss and feels the sting of being sidelined.
Meta’s Talent‑Buying Blitz
Zuckerberg has been spending heavily to reshape Meta’s AI division, poaching senior talent from Apple, OpenAI, Google and Anthropic with multi‑hundred‑million‑dollar packages.
These moves have upset many in the AI community, with OpenAI’s chief researcher publicly expressing frustration over the talent drain.
Yang Li‑kun’s Role at Meta
Since joining Meta in 2013, Yang led the core AI research projects, including open‑source frameworks, new agentic systems, and robotics‑AI integration.
In June, Meta spent $148 billion to acquire a 49 % stake in Scale AI, appointing its 28‑year‑old CEO Alexander Wang as chief AI officer, effectively placing a new head over Yang’s FAIR team.
FAIR’s Rise and Fall
Yang founded FAIR, which under his leadership produced breakthroughs such as ResNet (2015) and PyTorch (2016), and achieved major advances in computer vision and NLP.
However, Zuckerberg’s push for rapid commercialisation conflicted with Yang’s preference for incremental, open research.
Strategic Conflict Over Model Direction
Yang advocates building “world models” that understand physical reality, while Meta’s new Super‑Intelligent Lab focuses on large language models (LLMs) like GPT‑4, which are largely closed‑source.
Yang remains skeptical of LLMs, arguing they are merely statistical approximations lacking true understanding.
Llama 4’s Disappointment
Meta’s release of Llama 4, a 2‑trillion‑parameter model, received harsh criticism for underperforming in coding ability and context length, leading many engineers to abandon the model.
Rumours suggest internal disagreement over research philosophy contributed to the model’s failure.
Yang’s Vision vs. Business Reality
Yang’s recent work on the JEPA architecture aims to create models that grasp physical laws, but this long‑term research clashes with Zuckerberg’s demand for immediate commercial impact.
Meta’s shift toward product‑focused engineers and product managers reflects a preference for short‑term gains over foundational research.
Broader Implications
The story highlights the tension between academic excellence and commercial pressures in AI, echoing similar dilemmas faced by other pioneers such as Geoffrey Hinton and Demis Hassabis.
Yang’s experience underscores how even the most celebrated scientists can become “cogs” when corporate goals diverge from research ideals.
DataFunTalk
Dedicated to sharing and discussing big data and AI technology applications, aiming to empower a million data scientists. Regularly hosts live tech talks and curates articles on big data, recommendation/search algorithms, advertising algorithms, NLP, intelligent risk control, autonomous driving, and machine learning/deep learning.
How this landed with the community
Was this worth your time?
0 Comments
Thoughtful readers leave field notes, pushback, and hard-won operational detail here.
