The AI talent war just got another major casualty. Meta has successfully poached Yang Song, who led OpenAI's strategic explorations team, to become research principal of Meta Superintelligence Labs. Song started earlier this month, marking the latest high-profile defection in Silicon Valley's fierce competition for AI expertise.
Meta just landed another major win in Silicon Valley's AI talent wars. Yang Song, who spent three years leading OpenAI's strategic explorations team, has jumped ship to become research principal at Meta Superintelligence Labs, according to multiple sources reporting to WIRED. He's now reporting to Shengjia Zhao, another OpenAI alum who's been running Meta's buzzy superintelligence effort since July.
The move caps off what's been an aggressive summer hiring spree for Mark Zuckerberg. Meta has now poached at least 11 top researchers from OpenAI, Google, and Anthropic since June, signaling just how serious the company is about winning the race to artificial general intelligence.
Song brings some serious AI credentials to Meta's growing research empire. While still a Stanford PhD student, he developed breakthrough techniques that helped inform the development of OpenAI's DALL-E 2 image generation model. At OpenAI, his research focused on improving models' ability to process large, complex datasets across different modalities - exactly the kind of multimodal AI that's becoming crucial for the next generation of AI systems.
But Song's hire also highlights the complex web of relationships driving AI talent movement. Both he and his new boss Zhao attended Tsinghua University in Beijing as undergraduates, then worked under the same Stanford advisor, Stefano Ermon, while pursuing their PhDs. It's a reminder that in AI's tight-knit research community, personal relationships often matter as much as corporate recruiting budgets.
The timing couldn't be more significant. Zhao himself nearly returned to OpenAI earlier this summer, even going as far as signing employment documents before Meta convinced him to stay by formalizing his role as chief scientist of Meta Superintelligence Labs. In a July Threads post, Zuckerberg revealed that while Zhao had "cofounded the lab" and "been our lead scientist from day one," the company decided to make his leadership official.
Not everyone's staying put, though. The revolving door swings both ways in this talent battle. Two researchers have already returned to OpenAI from Meta Superintelligence Labs since June, with one going through Meta's entire onboarding process before never showing up for their first day. Another AI researcher, Aurko Roy, left Meta in July after just five months to join Microsoft AI, according to his personal website.
Song's arrival adds another layer of complexity to Meta's increasingly crowded AI hierarchy. When Zhao was promoted in July, some speculated he might be replacing Yann LeCun, Meta's longstanding chief AI scientist. LeCun quickly clarified in a LinkedIn post that he remains chief AI scientist for Facebook AI Research (FAIR), the company's original foundational AI research lab.
The talent shuffle reveals how Meta is essentially running parallel AI efforts - FAIR for fundamental research and Meta Superintelligence Labs for the race to AGI. It's a structure that mirrors how Google operates DeepMind alongside Google Research, though with the added pressure of trying to catch up to OpenAI's current lead in large language models.
For OpenAI, losing Song represents another hit to their research bench just as competition intensifies. The company has seen a steady stream of departures this year, from co-founder Ilya Sutskever to safety researcher Jan Leike. Each departure hands competitors like Meta valuable institutional knowledge about what's worked - and what hasn't - in pushing toward artificial general intelligence.
The broader implications extend beyond just these two companies. As AI talent becomes increasingly concentrated among a handful of tech giants, the moves signal how seriously Silicon Valley is taking the superintelligence race. Song's switch from OpenAI to Meta isn't just about individual career choices - it's about which company will have the research firepower to potentially crack the code on AGI first.
Yang Song's move from OpenAI to Meta represents more than just another executive shuffle - it's a strategic play in Silicon Valley's highest-stakes game. As companies race toward artificial general intelligence, the researchers who understand how to build these systems become the most valuable assets in tech. With Meta now housing talent from across the AI landscape under one roof, Zuckerberg is betting that concentrated expertise, not just computing power, will determine who reaches superintelligence first. The question isn't whether more defections will follow, but which company will emerge with the critical mass of talent needed to make the breakthrough.
