By Dow Jones Business News,
May 16, 2014, 02:33:00 PM EDT
By Paul Mozur and Rolfe Winkler
SUNNYVALE--Chinese search giant Baidu Inc. is often known as the Google of China. Now, it is setting up shop in
Baidu said Friday it will invest $300 million in a new research-and-development center in Silicon Valley that will
have almost 200 employees and be led by Andrew Ng, most recently the head of Stanford University's artificial-
intelligence lab. Mr. Ng, who also helped Google Inc. set up its artificial-intelligence efforts, was part of the
research team that built a computer that taught itself to recognize cats after looking at millions of YouTube videos.
The move highlights the growing ambition and influence of China's largest Internet companies. "There's been this
old stereotype that other countries copy U.S. technology," Mr. Ng said in an interview, "[but] there's tons of stuff
that Baidu has done that just isn't anywhere else in the world."
China's Internet companies have grown to be some of the largest Internet companies in the world, in part because
the Chinese government blocks foreign-based services such as Facebook and Twitter. The Chinese companies' users are
still mostly located in China, but they are increasingly active in Silicon Valley through investments and recruitment
Baidu rivals Alibaba Group Holding Ltd. and Tencent Holdings Ltd. both have offices near Silicon Valley that
spearheaded growing efforts to acquire and invest in local startups. Alibaba has filed for an initial public offering in
the U.S. that might become the largest listing in history.
Last year, Chinese smartphone startup Xiaomi Inc. hired executive Hugo Barra away from Google's Android mobile-
Baidu's new lab will focus on the development of artificial intelligence and deep learning, a field that has
recently drawn investment from the likes of Google, Facebook Inc. and Microsoft Corp.
Deep learning is a branch of artificial intelligence, in which computer scientists attempt to simulate the human
brain through computers that can see and learn for themselves.
Thus far, Internet companies have primarily used artificial intelligence to improve computer recognition of images
and speech, though deep learning could be used in an array of future technologies from driverless cars to robotics.
Baidu, which already has a deep learning center in Beijing, has used the technology to help refine how it targets
ads on its site and develop its speech-recognition software. Last year, it released an application that identifies
objects in smartphone photographs. It is also working on technology to predict tourist turnout at hotspots in China
based on weather, date, and other factors.
Mr. Ng, part of a small fraternity of researchers with expertise in deep learning, said the new Baidu lab would
give him more resources than he had at Stanford and the ability to assemble larger teams that could focus on specific
"The philosophy of the lab will be to have a small number of projects with strong leaders," said Mr. Ng. Adam
Coates, another Stanford artificial-intelligence researcher, is joining him at Baidu.
Baidu doesn't compete directly with Facebook and Google for users--since both face significant blocks in China--but
the three companies compete for talent.
Both U.S. companies have recently made big moves in a field that has a limited number of accomplished researchers.
Another well-known figure is Geoffrey Hinton, formerly of the University of Toronto, whom Google scooped up when it
bought his startup DNNresearch Inc. in early 2013 for an undisclosed amount.
Earlier this year, Google bought DeepMind Technologies for over $500 million, according to two people with
knowledge of the matter. DeepMind was a London-based startup that, among other advances, built a computer that learned
to master basic Atari videogames like "Pong."
Facebook late last year hired prominent New York University researcher Yann LeCun to lead its own artificial-
intelligence efforts. Microsoft has said it is using deep-learning technology to improve image searches for its Bing
Despite all the attention, computer brains remain primitive. Another leading academic in the field, Yohsua Bengio,
of the University of Montreal, said researchers are still able to build only "very tiny brains, like an insect brain."
Mr. Ng acknowledged the limitations, but said advances are happening quickly.
"There is a hypothesis that a lot of human intelligence is due to one learning algorithm," he said. "No one knows
what the right algorithm is, but it gives us hope that if we can discover some crude approximation of whatever this
algorithm is and implement it on a computer, that can help us make a lot of progress."
Write to Paul Mozur at firstname.lastname@example.org and Rolfe Winkler at email@example.com
Subscribe to WSJ: http://online.wsj.com?mod=djnwires
(END) Dow Jones Newswires
Copyright (c) 2014 Dow Jones & Company, Inc.