Baidu's Silicon Valley R&D center targets deep learning

Artificial intelligence is a new battleground for tech giants, and China's Web search leader Baidu, often termed "China's Google," is getting in on the action via a new research lab in Sunnyvale, Calif. In addition, the company announced it has hired Stanford computer science professor Andrew Ng, who previously worked on deep learning at Google (NASDAQ: GOOG), to head Baidu Research, which also includes labs in Beijing.

Baidu said it will invest $300 million in the Silicon Valley lab over five years.

Ng worked with Google researchers on the "Google Brain" project, which built a large-scale neural network that taught itself to identify images of cats after looking at millions of YouTube videos. Ng, who also co-founded Coursera, which offers online courses from top universities, will retain his position as chairman of Coursera's board.

Baidu launched its Beijing Deep Learning Lab, formerly called the Institute of Deep Learning, during 2013. The company claims to have made great strides in image recognition and image-based search, voice recognition, natural language processing and semantic intelligence, machine translation and advertising matching. In 2013, Baidu introduced an application based on deep learning that identifies objects in smartphone photographs.

Kai Yu, director of the Deep Learning Lab, told MIT Technology Review that Baidu's new Silicon Valley lab will focus on fundamental research, while scientists at the Deep Learning Lab will target the application of deep learning to new and existing Baidu products. "In Silicon Valley there's a huge talent pool that is so unique," said Yu. "We really want something revolutionary to come from the lab."

Though people are growing increasingly excited about the prospect for deep learning and neural networking, others say recent breakthroughs are more evolutionary than revolutionary.

Michael Mozer, a professor at the University of Colorado in Boulder and a board member of the Neural Information Processing Systems (NIPS) Foundation, told MIT Technology Review that core algorithms used by neural networks are much the same as those used in artificial intelligence during the late 1980s. Though these algorithms have been tweaked to work on a much larger scale, Mozer indicated that deep learning is not as dramatic of a breakthrough as some suggest.

Nonetheless, companies such as Facebook (NASDAQ: FB) and Microsoft (NASDAQ: MSFT) are hurriedly investing in deep learning technologies. Facebook hired New York University researcher Yann LeCun in late 2013 to work on artificial-intelligence efforts. And Microsoft's Bing search engine has benefited from the use of deep-learning technology to improve image searches. The company has also used deep learning to improve speech recognition and translation, as has Google.

In addition, Qualcomm (NASDAQ:QCOM) is deeply involved in investigating neural networking. Its Zeroth neuromorphic chip program is focused on signing up researchers this year to test its technology as the company eyes the commercial introduction of brain-mimicking chips as soon as 2015. The intent is to take neuromorphic computing platforms beyond mere object recognition to enable devices, such as smartphones, to take action based on information they have learned.

Google, meanwhile, has deepened its deep learning and artificial intelligence efforts by buying DNNresearch more than a year ago and paying in excess of $400 million for DeepMind Technologies earlier this year.

For more:
- see this Baidu release
- see this MIT Technology Review article
- see this Wall Street Journal article (sub. req.)

Related articles:
Qualcomm demos robot with neuromorphic chip
TeraDeep hopes to see Apple, Qualcomm, others use its artificial intelligence technology
Qualcomm, Ericsson and AT&T execs tout wearables, neuromorphic computing and more
Brain chips offer a new approach to crunching data