Quoc V. Le
Lê Viết Quốc, or in romanized form Quoc Viet Le, is a Vietnamese-American computer scientist and a machine learning pioneer at Google Brain, which he established with colleagues from Google. He co-invented the doc2vec and seq2seq models in natural language processing. Le also initiated and lead the AutoML initiative at Google Brain, including the proposal of neural architecture search.
Education and career
Le was born in Hương Thủy in the Thừa Thiên Huế province of Vietnam. He attended Quốc Học Huế High School before moving to Australia in 2004 to pursue a Bachelor’s degree at the Australian National University. During his undergraduate studies, he worked with on Kernel method in machine learning. In 2007, Le moved to the United States to pursue graduate studies in computer science at Stanford University, where his PhD advisor was Andrew Ng.In 2011, Le became a founding member of Google Brain along with his then advisor Andrew Ng, Google Fellow Jeff Dean, and researcher Greg Corrado. He led Google Brain’s first major breakthrough: a deep learning algorithm trained on 16,000 CPU cores, which learned to recognize cats by watching YouTube videos—without being explicitly taught the concept of a "cat."
In 2014, Le co-proposed two influential models in machine learning. Together with Ilya Sutskever, Oriol Vinyals, he introduced the seq2seq model for machine translation, a foundational technique in natural language processing. In the same year, in collaboration with Tomáš Mikolov, Le developed the doc2vec model for representation learning of documents. Le was also a key contributor of Google Neural Machine Translation system.
In 2017, Le initiated and led the AutoML project at Google Brain, pioneering the use of neural architecture search. This project significantly advanced automated machine learning.
In 2020, Le contributed to the development of Meena, later renamed LaMDA, a conversational large language model based on the seq2seq architecture. In 2022, Le and coauthors published chain-of-thought prompting, a method that enhances the reasoning capabilities of large language models.