According to Dynamic Beating Monitor, Google AI luminary and Gemini Model Technical Lead Noam Shazeer has once again left Google to officially join competitor OpenAI. OpenAI announced to its staff this Wednesday that Shazeer will focus on exploring a brand-new large-scale model architecture and driving the evolution of the Transformer architecture.
Shazeer was one of the co-authors of Google's seminal 2017 paper "Attention Is All You Need," which introduced the Transformer architecture that underpins modern generative AI models such as ChatGPT, Claude, and Gemini. He joined Google in 2000 and was one of Google's earliest core engineers. In 2021, due to Google's concerns about security and branding, they refused to release Meena, a chatbot developed by Shazeer and Daniel De Freitas. As a result, Shazeer chose to resign and co-founded Character.AI.
In August 2024, Google struck a deal with Character.AI worth a staggering $2.7 billion to rehire Shazeer. As the founder holding approximately 30% to 40% of Character.AI's shares, Shazeer personally cashed out between $750 million and $1 billion from the deal. Upon his return to Google, he served as Vice President and Co-Technical Lead of Gemini, primarily responsible for model pre-training. However, less than two years after Google spent a fortune to reacquire him, Shazeer has once again chosen to depart.
Shazeer's joining OpenAI is a significant gain for OpenAI in the talent competition, and it has caused the Google Gemini team to lose its core technical lead. Currently, large-scale models are reaching the compute limits of the Transformer architecture. OpenAI has tasked the original designer of Transformer to lead the team in finding alternative solutions, aiming to take the lead in the next generation of large-scale model architecture to compete against Anthropic and Google.
