Jeff Dean: Deep Dabbler

First Person: A conversation with Jeff Dean, senior fellow at Google Research

by Lisa Van Dusen / Palo Alto Weekly / Palo Alto Online

Try Googling "Jeff Dean jokes" and you will find many results — they seem to have a life of their own. While both flattered and a little embarrassed by this April Fools joke that keeps on giving, Dean is actually quite serious about his work. As a senior fellow at Google Research, Dean spends his days working on deep neural networks (also known as "deep learning," a subset of machine learning), which is modeled after the human brain. Artificial intelligence, deep learning and deep neural networks can all be applied in more ways than one can imagine.

For example, Dean's affinity for cats comes in handy with his line of work. In this context, cats are a mere vehicle for determining how much a computer can see, learn, communicate and understand. It also turns out that machines and humans are complementary in skills. While some computers are capable of beating a human opponent in a game such as Go, it's challenging for the same computers to perform more interpretive functions such as identifying and describing images. On the other hand, humans (and cats) are challenged by performing algorithmic functions on large sets of data, a task at that machines excel at. Dean is currently working on helping computers get better at "reading" things, including images, sounds, games and more.

 Jeff Dean's special Google ID Badge for the Deep Neural Networks Team, complete with steel cable and brain.

Jeff Dean's special Google ID Badge for the Deep Neural Networks Team, complete with steel cable and brain.

So why does this matter? Dean believes that his work could help address many of the National Academy of Engineering Grand Challenges — important national or global problems believed to be solvable through science and technology such as engineering better medicines or providing clean access to water.

Dean is thirsty to utilize the tools he has created through deep neural networks in service of real world scientific applications.

Back in 1999, not long after earning his doctorate degree in computer science at the University of Washington, Dean scrambled to prepare Google's on-site servers for Monday morning-traffic surges. Still inspired by collaborating in small teams, Dean is never tired of the challenges and opportunities at Google.

A self-described "dabbler," Dean is eager to read a wider range of non-fiction in order to learn about different scientific fields. But at the end of the day, he just hopes people see him as a nice guy.

 

Host/interview, Lisa Van Dusen

Video, Veronica Weber

Production Manager, Lavanya Mahadevan