Yesterday (February 24), Baidu released the second quarter of 2016 earnings and 2016 full-year earnings, earnings report, 2016 annual revenue of 70.549 billion yuan, an increase of 6.3% over 2015. Baidu 2016 annual net profit of 11.632 billion, far lower than 2015 to achieve 33.6 billion yuan, accounting for only about one-third of its. The Lei Feng network found in Baidu's earnings, the R & D spending this one to invest 10.15 billion yuan, the basic and net profit was flat. This huge R & D investment, and Baidu repeatedly referred to the "artificial intelligence" vocabulary has a great relationship.
Everything around artificial intelligence
In Baidu's earnings statement, Li said:
"Our existing platform products, including search engines and information streams, have been optimized for upgrading through artificial intelligence technology applications and are further enriched by our content and service ecosystems. We are excited about the tremendous opportunities brought about by the opening of the existing platform, and we will actively build a new era of artificial intelligence-driven business in the future. & Rdquo;
Baidu CFO Li Xin also said:
As in the past, we will create the best experience for users while investing in AI-driven areas such as finance, cloud and autopilot. & Rdquo;
In the slowdown in growth and decline in net profit in front of the facts, Baidu with artificial intelligence for their own sketched out a still possible future. Baidu also disclosed further personnel changes: Lu Qi will serve as Baidu company directors and vice chairman of the board; since July 2004 has served as a member of the board of directors of the Greg & middot; Pana (Greg Penner) from the board and Chairman of the board of directors of the board of directors; Lenovo Group Chairman Yang Yuanqing, was appointed Baidu Board of Directors under the Audit Committee members and corporate governance and nomination committee members.
Baidu said in the earnings report, the fourth quarter of major personnel adjustments mainly for the new artificial intelligence for the new channel to speed up the overall upgrade transition.
Everything is around artificial intelligence.
R & D costs: 2 years 20 billion
Baidu 2016 R & D cost is very large: the total cost of 10.15 billion yuan, accounting for 14.3% of total revenue. Such a large proportion, in the global high-tech companies are relatively rare.
A few years ago, Baidu's research and development costs began to soar, starting from 2014, Baidu's R & D costs began to surge.
In 2014, R & D expenses were 6.981 billion yuan, an increase of 70% over 2013,
2015, R & D investment rose to 10.176 billion yuan, an increase of 45.8% over 2014,
2016, R & D costs of 10.15 billion yuan, 0.2% less than 2015
We found that R & D investment soared from 2014, and in 2014 Baidu's most important event is to increase the intensity of the three R & D center talent.
January 2014, the depth of the Institute of Learning (IDL) began to launch "young Marshal plan" for 30 years of outstanding artificial intelligence technology talent selection and training. The first batch of "young Marshal plan" with nine places, selected "young Marshal plan" staff can directly get Baidu T9 to T10 rank, annual salary of millions of yuan start, not cap.
May 2014, the world's top artificial intelligence expert Wu Enda out of Baidu chief scientist, a comprehensive leadership Baidu Research Institute, also announced the Silicon Valley AI Laboratory (SVAIL) was formally established.
In July 2014, Baidu set up a large data laboratory (BDL), in Beijing, Shenzhen and Silicon Valley office. The laboratory is mainly engaged in large data and search core technology-related research work, focusing on large-scale machine learning algorithms and applications, large data forecast analysis and vertical industry application exploration, large data structure algorithm research, intelligent system research and so on.
It can be said that Baidu R & D costs to the extent of today, with the preparation of these three major laboratory closely In response to the crazy growth of R & D investment in the past two years, the official reasons are also very direct:
This growth is mainly due to the increase in the number of R & D personnel & rdquo ;.
Therefore, Baidu's annual research and development of more than 100 billion investment, mainly spent in the preparation and maintenance of talent pool. 2015 and 2016 two years, Baidu R & D investment in 10 billion yuan or so floating, indicating that by the end of 2015, Baidu's R & D team recruitment has been basically completed,To Baidu's financial and physical, to build a complete set of AI talent pool spent two yearsThe
Two R & D center staff
According to official website information, SVAIL laboratory about 50 people. However, this number is not a lot of R & D team, hired Silicon Valley first-class computer, mathematics, physics and many other areas of top scientists. Most of them come from Stanford University, UC Berkeley and Carnegie Mellon University and other American schools, which is no lack of professors. In addition, there are a lot of people are from the NVIDIA, Google such a first-class North American companies hired over, so Baidu in which the talent into the show.
We can look at the following representatives.
The direct leader of the SVAIL lab is Adam Coates, who will be selected by the "MIT Technology Review" in 2015, 35 under the age of 35. Stanford University Ph.D. in computer science, the main research direction for the automatic driving computer vision, depth learning HPC method, depth learning and large-scale machine learning and other fields.
Patrick LeGresley (research scientist), joined Baidu since 2015. Dr. Stanford Aerospace, before Baidu, he served as a lecturer at the Stanford University Computing Mathematics Engineering (ICME) for three years. The main direction for all aspects of HPC, including algorithms, processors, high-performance networks and storage.
Sharan Narang (Software Engineer), joined Baidu in October 2015 so far. He worked for more than six years at NVIDIA and served as technical director for Tegra SoC. He recently received two papers by ICLR2017.
Sanjeev Satheesh (research scientist), from August 2014 to the present. Graduated from Stanford University, Master, Stanford University CV team assistant researcher (nearly 2 years), when he was in the Imagenet team. Then worked as a development engineer at Microsoft for two years. And then came to Baidu.
Ryan Prenger (research scientist), in August 2014 to join Baidu so far. Doctor of Physics, University of California, Berkeley, has five patents.
Andrew Gibiansky (Software Engineer and Machine Learning Scientist), he recently promoted the introduction of in-depth learning of the Ring allreduce algorithm in the HPC field.
Some other researchers also include:
Gregory Diamos (Research Scientist) joined Baidu in August 2014. Main attack HPC training depth learning neural network. Prior to working at NVIDIA for nearly two and a half years, CUDA Software Architect.
Christopher Fougner (research scientist), joined Baidu in April 2015. Stanford University Master of Mathematics in Mathematical Engineering, the main direction of speech recognition.
Carl Case (research scientist), from September 2014 to join Baidu so far. Master of Computer Science, Stanford University.
Eric Battenberg (Research Scientist), joined Baidu in April 2015. UC Berkeley Ph.D. in Electrical Engineering and Computer Science.
Anuroop Sriram (research scientist), joined Baidu since 2015. Previously worked as a data scientist at Twitter, Master of Computer Science at Carnegie Mellon University.
Jared Casper (Senior Software Engineer), joined Baidu since 2014. Stanford University Ph.D. in computer science, previously served as senior software engineer at Oracle Labs.
About June 2016, one after another wave of R & D personnel to leave the SVAIL laboratory, in the Lei Feng network (public: Lei Feng network) to find the information, these people often flow to Google, OpenAI, NVIDIA and Amazon AI R & D department , And often in the new company to obtain a more senior position, we can see Baidu SVAIL laboratory experience to their resume to increase the considerable weight. The representatives of these departures are:
Bryan Catanzaro: June 2014 to join Baidu. He led a team of 15 research and development teams in Baidu to explore the field of speech recognition, training and deployment of deep neural network tools and methods. Prior to Baidu, he took office in the British Weida for 3 years, and in September 2016 to leave Baidu, re-joined the British Weida, as the British Weida depth study and application research department vice president.
Erich Elsen: joined Baidu in September 2014. Was an associate professor at Stanford University, worked in Baidu for 2 years, in September 2016 to switch to Google, any Google brain senior software engineer, focused on Tensorflow and enhanced learning.
Jesse Engel, joined Baidu in January 2015. UC Ph.D. in materials science from the University of Berkeley, left in July 2016, joining the Google Brain as a research scientist.
Dario Amodei, joined Baidu in September 2014 and left after a year. Later joined Google as a senior research scientist, joined OpenAI in July 2016 as a research scientist.
Rishita Anubhai, joined Baidu in 2015 as a machine learning engineer. In September 2016 joined the Amazon depth learning department.
We see these researchers repeatedly in Google, NVIDIA, Amazon, Baidu and other companies between the flow, reflecting the North American AI is now fierce competition for talent situation. And Baidu as a Chinese company, to maintain the attractiveness of North American talent, to maintain their own talent pool and attract new talent to join, which to invest in the cost of competition with the talent in the country is quite different.
IDL Research Institute
In addition to Wu Enda, IDL Research Institute is the main person in charge of Lin Yuanqing, senior director of IDL. Lin received his Ph.D. in electrical engineering from the University of Pennsylvania in 2008. He joined NEC Labs America as a researcher to study feature learning and mass classification. In 2010, he led the NEC-UIUC team to be the first in the ImageNet Large-Scale Visual Challenge. Since April 2012, he has been the head of the NEC Labs America media analysis department, focusing on two research directions: 1) large-scale fine-grained image recognition based on mobile clouds, and 2) 3D visual perception of autopilot. In November 2015, he joined Baidu and served as senior director of the Depth Learning Laboratory. His research direction covers machine learning and computer vision. He also served as president of the NIPS conference in 2015.
IDL Deputy Director Wu Zhongqin: Wu Zhongqin the master are studying at the Fudan University Department of Computer Science, in July 2008 after graduation to join Baidu. In Baidu 8 years, he has responsible for the text, voice, image technology R & D team management, had the team won the company's best team, for two consecutive years won the company's highest award. During his tenure at the depth of the lab, he led Baidu's image technology team to achieve leapfrog development, with good technical influence and promoted vice president in 2015. In 2016 he built the AR team from scratch, and now is committed to Baidu AR and Baidu brain research and development.
IDL "outstanding scientist" Xu Wei: responsible for the development of the depth of learning platform and algorithm research. Xu Wei received a bachelor's degree in computer science from Tsinghua University in 1998 and a master's degree from Carnegie Mellon University in 2000. Prior to joining Baidu, Xu Wei worked for the US Facebook, created and responsible for large-scale recommendation platform, and widely used in a number of products for the billions of users to provide personalized service. Prior to Facebook, Xu Wei served as a senior researcher at NEC Lab, where depth learning techniques were applied to image recognition products.
In addition to the above three leaders, Baidu IDL Institute of public information show that nearly 90 researchers. There are several major research projects: product image retrieval, Baidu map, autopilot, three-dimensional computer vision, ImageQA and other fields.
Summary: Lei Feng found that SVAIL laboratory seems to be a new wave of large-scale recruitment work. In LinkedIn, many SVAIL researchers have been in their own home page posted Baidu's job advertisements. Baidu today released the report shows that it is facing a slowdown in growth and profit decline in the harsh facts, while Baidu is still continuing to promote in the field of AI, especially in the field of basic research and development investment, for a very mature business company, In the artificial intelligence of the arrival of the moment, in research and development to make such a big investment, the AI industry has played a very good example and reference.