Jeff Dean
Chief Scientist
Google DeepMind
Technical architect of Google's AI infrastructure, including TensorFlow and Google Brain. Now Chief Scientist at Google DeepMind. A key figure in scaling deep learning systems to production at global scale.
Core Positions & Ideas
Large-Scale Distributed Computing Is the Foundation of AI Progress
2012Co-developed Google's DistBelief and later TensorFlow — infrastructure that made training large neural networks at Google scale possible. His thesis: AI breakthroughs are partly about algorithms, but equally about having the distributed computing systems to train larger models than anyone thought practical.
Multi-Task Learning and Transfer Learning Are Underrated
2017Championed the idea that a single model trained on many tasks simultaneously often outperforms specialized single-task models — a precursor to the 'generalist AI' that foundation models represent. Pathways (2022) extended this to a single model learning thousands of tasks.
AI for Health and Science Is the Most Important Application
2021Leads Google DeepMind's efforts in scientific and health AI. Argues that AI's most valuable contribution will not be in consumer products but in accelerating scientific discovery: disease diagnosis, drug design, climate modeling, and materials science. AlphaFold is the existence proof.
Essential Reading & Watching
TensorFlow: A System for Large-Scale Machine Learning
The TensorFlow paper. TensorFlow became the dominant deep learning framework for several years and remains widely used. Co-developed by Dean's team at Google Brain.
Read / Watch →
A Decade of Deep Learning: Progress, Challenges, and Opportunities
Dean's retrospective on a decade of deep learning progress at Google. A detailed technical survey of what worked, what didn't, and what remains unsolved.
Read / Watch →