Noam Shazeer

Noam Shazeer

A co-author of the 2017 Transformer paper and one of the most prolific architects of the machinery around it: sparsely-gated mixture-of-experts layers, which let a model activate only a fraction of its parameters per token; multi-query attention, which made inference far cheaper; and the T5 text-to-text framework. He left Google to co-found Character.AI, and returned in 2024 to co-lead the Gemini models. (See also: Transformer, Inference, Parameters, Ashish Vaswani)