Noam Shazeer
Noam Shazeer is an American AI researcher and engineer best known as one of the eight co-authors of the landmark 2017 paper Attention Is All You Need, which introduced the Transformer architecture that now underlies nearly every modern Large Language Model. He later co-founded Character.AI, one of the first companies to bring free-flowing, personality-driven conversational agents to a mass audience.
Shazeer joined Google in 2000, where his early contributions included improving the search engine's spelling corrector — a humble beginning for someone who would go on to help redefine deep learning. His research career threaded together many of the ideas that made large-scale machine learning models practical: he was a driving force behind Mixture of experts layers, which route computation through specialized sub-networks rather than activating an entire model at once, and he helped design distributed training systems like Mesh-TensorFlow for training models across supercomputers. His work pushed language modeling away from earlier paradigms built on recurrent neural networks and Long short-term memory, replacing sequential recurrence with parallel Self-attention — the conceptual leap at the heart of Attention Is All You Need alongside collaborators including Ashish Vaswani.
At Google, Shazeer and his colleague Daniel de Freitas built a chatbot named Meena. Following the refusal of Google to release the chatbot to the public, Shazeer and Freitas left the company in 2021 to found Character.AI, a platform letting users converse with customizable AI personas, blurring lines between chatbots and interactive fiction. In September 2023, Time Magazine chose Shazeer as one of the 100 most influential people in the AI world. In 2024, in an unusual arrangement, Google effectively absorbed Character.AI's technology and brought Shazeer back to co-lead its Gemini AI efforts.
Shazeer's career exemplifies the restless, iterative spirit of AI research — leaving, building, returning — always chasing the frontier of what Attention-based systems can do.