Chris Unger/Getty Images
Meta has heavily recruited from Google’s AI division DeepMind for its new superintelligence team, bringing in at least 10 previously unreported researchers since July, according to LinkedIn data reviewed by Business Insider.
Meta’s DeepMind hires include researchers who worked on Google’s most powerful AI models, and even one whose work helped Google earn gold at the International Mathematical Olympiad.
Meta has also hired heavily from Scale AI for its safety and evaluations team. At least six of the data labeling startup’s researchers announced their departures to Meta in recent weeks. In June, Meta bought nearly half of Scale AI for $14 billion, tapping its CEO Alexandr Wang to lead its superintelligence effort.
Meta’s recruiting campaign has made headlines for its eye-popping offers and for poaching staff from OpenAI, as the company competes against other Big Tech companies in the AI talent race. DeepMind’s strong presence at Meta Superintelligence Labs, or MSL, shows Google remains a powerhouse of AI research as well.
Meta has also invested in and hired from Scale’s research division, which has expertise in fields key to improving chatbots and making sure they don’t produce toxic or harmful content.
“We operate one of the most important research teams in AI, and we’re growing it. The moves of a small number of employees to Meta as part of the transaction are a distinct dynamic from the other companies mentioned in this story,” Joe Osborne, a Scale AI spokesperson, told Business Insider in a statement.
Meta declined to comment. Google did not respond to a request for comment from Business Insider. Business Insider also reached out to the researchers Meta hired from Google and Scale AI for comment.
Meta targets DeepMind
One of Meta’s most notable hires is Yuanzhong Xu, a veteran Google researcher who contributed to several of the search giant’s most prominent AI initiatives.
Xu, who previously held a senior engineering position at Google DeepMind, was instrumental in the development of LaMDA, the company’s conversational AI model family, and PaLM 2, a large language model that served as the foundation for Gemini, Google’s flagship family of generative AI models. Xu’s work also helped power Gemini’s development.
Xu’s LinkedIn profile says that he now works at “TBD” at Meta, a likely reference to a unit within Meta’s Superintelligence Labs dedicated to training the company’s most advanced AI models.
Meta also hired Tong He, whose work Google credited with helping it achieve gold medal-level performance at the International Mathematical Olympiad in July. Meta hired three other DeepMind researchers behind the win, The Information reported at the time.
Another former DeepMind researcher, Mingyang Zhang, who Meta hired in July, has worked on helping large language models retrieve information effectively.
Meta has also hired Xinyun Chen, a former DeepMind researcher whose expertise centers on AI systems capable of autonomous code generation and logical reasoning.
Google CEO Sundar Pichai has previously dismissed the impact of Silicon Valley’s raging talent wars, saying its retention remains “healthy.”
Meta hires from Scale AI
Scale AI has a SEAL (Safety, Evaluations, and Alignment Lab) team that oversees crucial areas in AI to help ensure chatbots align with human values, so they don’t do things like spew misinformation or instructions for building a bomb. Evaluations focus on determining an AI model’s performance and how to improve it.
From Scale’s SEAL, Meta recruited Ziwen Han and Nathaniel Li, who both were coauthors on a test for AI models called Humanity’s Last Exam. This test is meant to be the hardest possible test that humans can devise for an AI to take.
Other recent Meta hires from Scale include Alexander Fabbri, Felix Binder, Chen Xing, and Lifeng Jin. Many of them now work at MSL on safety and evaluation-related issues.
These aren’t the first Meta hires from Scale. Days after Meta unveiled its superintelligence team, SEAL’s first leader, Summer Yue, joined the company as part of Meta’s investment in Scale, according to her LinkedIn. She now leads the alignment group at MSL.
Julian Michael, who led SEAL after Yue, made the jump to Meta around the same time.
Have a tip? Contact Pranav Dixit via email at [email protected] or Signal at 1-408-905-9124. Contact Charles Rollet via email at [email protected]m or on Signal and WhatsApp at 628-282-2811. Use a personal email address, a nonwork WiFi network, and a nonwork device; here’s our guide to sharing information securely.
The post Meta raids Google DeepMind and Scale AI for its all-star superintelligence team appeared first on Business Insider.
