Google's New AI Head Is So Smart He Doesn't Need AI

Colleagues joke that Jeff Dean increased the speed of light. Now he's charged with taking Google's artificial-intelligence efforts in new directions.
Image may contain Jeff Dean Human Person Wood Teacher and Plywood
Jeff Dean, who joined Google in 1999 when the company was a startup, is now head of Google's artificial-intelligence efforts.Google

Google’s heavy investment in artificial intelligence has helped the company’s software write music and beat humans at complex board games. What unlikely feats could be next? The company’s new head of AI says he’d like to see Google move deeper into areas such as healthcare. He also warns that the company will face some tricky ethical questions over appropriate uses for AI as it expands its use of the technology.

The new AI boss at Google is Jeff Dean. The lean 50-year-old computer scientist joined the company in 1999, when it was a startup less than one year old. He earned a reputation as one of the industry’s most talented coders by helping Google become a computational powerhouse with new approaches to databases and large-scale data analysis. Google colleagues once created a joke website of “Jeff Dean facts,” including his purported role in accelerating the speed of light. Another had it that Dean doesn’t really exist—he’s an advanced AI created by Jeff Dean.

Dean helped ignite Silicon Valley’s AI boom when he joined Google’s secretive X lab in 2011 to investigate an approach to machine learning known as deep neural networks. The project produced software that learned to recognize cats on YouTube. Google went on to use deep neural networks to greatly improve the accuracy of its speech recognition service, and has since made the technique the heart of the company’s strategy for just about everything.

The cat-video project morphed into a research group called Google Brain, which Dean has led since 2012. He ascended to head the company’s AI efforts early this month after John Giannandrea left to lead Apple’s AI projects.

Dean’s new job puts him at the helm of perhaps the world’s foremost AI research operation. The group churns out research papers on topics such as creating more realistic synthetic voices and teaching robots to grasp objects.

But Dean and his group are also on the hook to invent the future of Google’s business. CEO Sundar Pichai describes the company’s strategy as “AI first,” saying that everything the company does will build on the technology.

One way Dean hopes to help is by conjuring up new lines of business. Google’s AI research has so far mostly been used to improve or expand existing products, such as search and smartphone software. “New machine learning capabilities or research results might enable us to do new things that Google doesn't now,” says Dean. “Health is one that's pretty far along in this direction.”

Dean wouldn’t discuss details. But two Google research projects offer clues. The company is testing software in India that can detect a complication of diabetes that causes blindness, and has also tested software that looks for signs of breast cancer on microscope slides. The FDA has begun cautiously approving AI software that helps doctors make medical decisions.

Success in healthcare could help diversify the business of a company that, despite broad interests, relies heavily on advertising. In 2017, almost 90 percent of parent company Alphabet’s revenue came from ads.

Dean is also excited about automating artificial intelligence---using machine-learning software to build machine-learning software. “It could be a huge enabler of the benefits of machine learning,” Dean says. “The current state is that machine learning expertise is relatively in short supply.”

Google dubs that meta-AI project AutoML. Dean says Alphabet’s self-driving car division Waymo has been testing the technology to improve its vision systems. AutoML is also at the core of a new cloud service of the same name that helps companies create customized image-recognition systems.

Longer term, Dean thinks automatic AI could enable robots to figure out how to cope with unfamiliar situations, such as opening a type of bottle cap they haven’t seen before. Despite successes on narrow tasks, AI researchers have struggled to make machines capable of many different thing. Dean says AutoML could bring about “really intelligent, adaptable systems that can tackle a new problem in the world when you don't have exposure of how to tackle it.”

More powerful AI systems will also lead Google into new, uncertain moral ground. Some researchers at the company are working on how to ensure that machine-learning systems don’t breach societal expectations of fairness. In 2015, Google’s photo-organizing product tagged some images of black people as gorillas, and the service has been blind to searches for the black apes ever since.

Google also faces other ethical decisions over allowing customers to tap its AI skills. The same week Dean took over Google’s top AI role, the New York Times reported that thousands of Google employees had signed a letter protesting a contract with the Pentagon that applies AI to interpret video from drones.

Dean declines to comment on the project’s details, but suggests it’s more a pointer to future ethical questions than an immediate moral challenge. “The actual project is relatively mundane, sort of assembling a bunch of existing open source components together, but it does I think give us pause as a company about what we want our role to be,” Dean says. “There's a wide range of views in the company about what we should be doing there.”

Google’s AI Adventures