Google buys startup company built on CIFAR research
When Google Inc. bought Program Director Geoffrey Hinton’s startup company DNNresearch last week, the search engine giant acquired both technology and talent that grew out of long-term collaborative research at the Canadian Institute for Advanced Research.
Hinton, who is now a Google employee as well as a University of Toronto professor, is one of the world’s leading experts on “deep neural networks,” a type of artificial intelligence that can, among other things, give search engines a raft of new capabilities.
Hinton has researched machine learning since the 1970s. CIFAR helped recruit him to Canada and U of T 25 years ago. Since 2004, Hinton has directed CIFAR’s Learning in Machines & Brains program (formerly known as Neural Computation & Adaptive Perception), a group that convenes experts in machine learning, computer vision, and neuroscience to collaborate on building better perceptual systems for machines and understanding the perceptual systems of the brain.
“The CIFAR program created a community of researchers who could push the development of ‘deep learning’ ahead farther and faster than any of us could have done on our own,” Hinton said. “Now, I am pleased to work with Google on ways to apply that knowledge so as to improve people’s lives.”
Deep neural networks are very good (and getting better) at voice and image recognition, both of which are growth areas for search engines.
“Suppose you are hiking and see a patch of mushrooms that you’re not sure are edible,” Hinton says. “Rather than trying to type a description of the fungi into a search engine, you can just snap a picture with your smartphone. A deep neural network can analyze the image itself and find information on that particular species.”
Deep neural networks also make voice searching much more accurate and efficient, and even improve conventional text searches by better analyzing the context of both queries and results.
For instance, deep learning could allow a machine to understand that the headlines, “Canada triumphs at national game,” and “United States’ Olympic hockey team trounced,” are related, even though they share no words in common.
Convinced that deep neural networks had reached the stage where they could find widespread commercial application, Hinton and two of his grad students created the company DNNresearch in 2012. This business acquisition suggests that Google’s strategists share the researchers’ conviction.
“We think Geoff and his team’s approach to problem-solving will be valuable to a variety of projects that we’re working on at Google,” a spokesperson for the company said. “The deep learning approach has already had a significant impact in the recognition quality of Voice Search.”
As society looks to the future and the potential of AI and robotics, CIFAR Senior Fellow and MIT economist Daron...