Home > News content

Overall layout AI, IBM, Google is how to do?

via:博客园     time:2016/11/9 21:31:25     readed:2756

This article excerpts from the Great Wall Securities Report - the Internet ushered in the AI ​​era, overseas technology giants first to layout: artificial intelligence depth report (foreign chapter one), without altering the original intention on the basis of a slight deletion.

PC era of the Internet core competitiveness of software products for rapid response capability, mobile Internet era is to build the mobile end of the ecosystem, artificial intelligence era is more dependent on the AI ​​core technology.

AI technology has two elements:

Core technology platform

Data cycle

Only the AI ​​technology and data combine to form a practical business. This paper focuses on IBM, Google in the basic level, technical level, the overall application layer layout AI, and its extended application scenarios and other content.

IBM & mdash; Watson

Artificial intelligence is IBM in 2014 after the focus areas, IBM in the AI ​​field layout around Watson and brain-like chip, trying to build AI ecosystem. At present, IBM has withdrawn global business consulting GBS and technical services GTS and other departments, and transformation into cognitive solutions and cloud platform company.

全面布局全面布局 AI,IBM、Google 是如何做的?

IBM's core strategy for the next decade is the "wisdom of the Earth" program, IBM R & D investment in its annual investment of about 30 billion dollars.

The future of IBM's innovative solutions in intelligent energy, intelligent transportation, intelligent medical, intelligent retail, intelligent energy and intelligent water resources in full blossom, covering energy saving, food safety, environmental protection, transportation, medical, modern services, software and Services, cloud computing, virtualization and other hot direction.

2016 Q3, Watson, represented by cognitive solutions services to achieve revenue of 12.889 billion US dollars, revenue growth is rapid, accounting for up to 22. 17%, IBM in the AI ​​field of profit began to erupt. We expect IBM Cognitive Solutions to achieve revenue of 19.039 billion yuan, 21.895 billion yuan and 24.084 billion yuan, respectively, when cognitive solutions will account for 24.56%, 26.89% and 28.72% of IBM's revenue. IBM's main business growth performance.

Watson Leads Cognitive Business:

At present, IBM no longer Watson as a single system business, and its function is divided into different components, each part can be rented out to solve a specific business problem. IBM Watson, represented by cognitive technology to business into the cognitive business era, to help businesses tap the commercial value, reshape the industrial structure. IBM is providing innovative solutions to customers, customers continue to import their own business data Watson and Watson training.

Cognitive Computing, Large Data Analysis, Internet of Things, Heterogeneous Computing, Synapse and Cognitive Machine Systems, etc., are gradually emerging into new energy applications, pollution Urban management, ecological improvement, health care, transportation, food safety traceability and community services.

Watson uses advanced natural language processing, information retrieval, knowledge representation, reasoning, and machine learning techniques to analyze large numbers of evidence, generate hypotheses, and analyze and estimate them when analyzing problems and determining the best solution. At present, Watson has developed 40 different products, including common language recognition services. Watson is good at cognition, designed for understanding, reasoning, and learning. He has the opportunity to overcome challenges that he has never met, such as the challenges of health care, water management challenges, insurance fraud challenges, fashion challenges, Environmental challenges, out-of-the-box M & A risk challenges.

Watson + Healthcare Build Healthcare Platform:

Watson focuses on the diagnosis of cancer and cancer in the field of medicine, which has the advantage of natural language processing, through the excavation of unstructured data to find deep relationships. Watson Medical's business strategy is:

1, the depth of focus on the field of cancer, and to other areas of expansion;

2, access to large-scale acquisition of data resources;

3, through cooperation and other expansion of the use of the scene, the output of ecological capacity.

With the convergence of healthcare data, people, capabilities, and customers, Watson Health will be a huge health care platform, with Watson Cognitive Computing powering the wisdom of healthcare. Watson efficiency, accuracy significantly higher than the human, "cognitive computing + medical" prospects, IBM profound benefit from the industry development dividend.

In addition, IBM, with its strong cognitive computing capabilities, applied to digital consultants, virtual assistants, cloud computing, scientific research and other fields, vigorously develop quantum computing circuit, open quantum computing platform, introduced a variety of parallel brain- Enhance AI arithmetic. November 2015, IBM open source artificial intelligence basic platform SystemML, can support descriptive analysis, classification, clustering, regression, matrix decomposition and survival analysis algorithms, Watson integrated a number of SystemML function.

Google & mdash; combination of hardware and software, open source system to build AI ecology

Google is the world's largest data retrieval core technology, and the establishment of the world's largest database system. Advertising profit is Google's main profit model, the current revenue of more than Jiucheng from its advertising system.

August 2015, Google announced the restructuring, the establishment of parent company Alphabet, Google search engine company by the comprehensive shift to cover many areas of high-tech enterprises.

全面布局全面布局 AI,IBM、Google 是如何做的?

Google has set up AI in 2011, and now has more than 100 teams using machine learning technology, including Google search, Google Now, Gmail, etc., and its open source Android mobile phone system into a lot of machine learning Neural Network to Develop Android Phone Speech Recognition System). Google currently relies on major AI-driven products and services, such as Google's use of in-depth learning technology to improve search engines, identify Android phone instructions, and identify the images of their Google+ social networks.

Google's approach to developing AI is:

1, covering more users to use the scene, from the Internet, mobile Internet and other traditional business extends to the smart home, automatic driving, robots and other fields, the accumulation of more data and information;

2, the accumulation of the underlying artificial intelligence technology, research and development of more advanced depth learning algorithm to enhance the pattern recognition and voice recognition capabilities, the information for deeper processing and processing. Google tried to penetrate the AI ​​to its products, to bring more users to use the scene, and more intelligent features.

全面布局全面布局 AI,IBM、Google 是如何做的?


全面布局全面布局 AI,IBM、Google 是如何做的?

November 2015 Google open source second - generation depth learning system Tensorflow. Tensorflow can write and compile code for executing machine learning algorithms, and turn machine learning algorithms into symbolic representation of various types of graphs, shortening the time to rewrite the code. TensorFlow mimics the way the human brain works and recognizes patterns that are used in many areas, such as speech recognition or photo identification. In addition, the TensorFlow-written operations can be run on a wide variety of heterogeneous systems with little or no change. In the open source, all engineers will help Google to modify and improve the technology, Google received feedback, can be introduced to better services and products, and thus promote the development of the entire AI industry.


DeepMind was founded in 2010, which combines machine learning with the most advanced technology of systems neuroscience to build a powerful general-purpose machine learning algorithm.

In January 2014, Google spent $ 263 million acquisition of Deepmind.

In December 2014, Google through DeepMind and Oxford University's two AI research team established a cooperative relationship.

February 2015, Deepmind system learned 49 Atari classic game.

In March 2016, DeepMind developed AlphaGo to beat world go champion Li Shi-shih with a 4: 1 victory to inspire worldwide attention to artificial intelligence.

The current AlphaGo focus on the development of chess, the future will also be used in medical diagnosis, or into the unmanned field, to accelerate the AI ​​commercial process.

Virtual assistant integrates intelligent home, promote ecological construction

Google believes that the field of smart home applications will be the future of AI is an important market, the world's smart home penetration is low, for this Google is accelerating to Nest, Google Assistant smart home based ecosystem building, through a series of mergers and acquisitions, Open platform, software and hardware integration to build the ecosystem.

Google launched in May 2016 voice intelligence assistant Google Assistant, is the voice recognition, artificial intelligence, natural voice comprehension master.

Google Assistant can fully understand the context and answer questions, and Alexa, Siri and Hound and other intelligent assistants to compete. Compared to Google Now is mainly used for mobile phones and PC, Google Assistant is beginning to integrate into a variety of devices (Google Home, Allo chat robot). According to Markets and Markets, the natural language processing market will grow from $ 7.63 billion in 2016 to $ 16.07 billion in 2021, with annual growth of 16.1 percent.

In June 2014, Google spent $ 555 million acquisition of cloud-based home control company Dropcam, October and acquisition of intelligent home central control equipment company Revolv, Revolv will participate in Nest "Works with Nest" open plan. In May 2016, Google Home (Smart Speaker) was introduced. Google Home is a smart speaker based on Google Assistant voice controls. Compared to the Amazon Echo, Google Home will use Google's huge database to understand user needs.

Sensor unifies the AI ​​algorithm research and development unmanned pilot prototype vehicle

Google unmanned vehicle project began in 2009, 2011 for its acquisition of 510 Systems, Anthony's Robots and other companies. At present unmanned driving mileage of 1.8 million miles, and successfully released the world's first fully capable of automatic pilot prototype car & rdquo; & rdquo;, and claims that by 2020 Google Auto will be officially listed.

Google unmanned technology-driven, focusing on basic technology research and AI core technology development. In the capture of relevant depth learning and brain technology development and other software algorithms based on the integration of a variety of sensors. In December 2015, Google and Ford will set up a joint venture company, based on Google AI technology research and development of unmanned vehicles, vehicle technology can save time and money.

Jointly NASA research and development of quantum hardware, released TPU into the chip market

Google has established a quantum artificial intelligence laboratory (Quail), the laboratory by the United States of America NASA (NASA), University Space Research Association co-host. In 2013, Google has been using D-Wave machines in the Web search, voice / image pattern recognition, planning and scheduling, air traffic management, robot space missions and other applications to explore quantum computing, and support the task control center operation. In 2014, Google used its experience in the D-Wave machine to develop quantum hardware, through the appointment of the University of California professor of physics John Martinis and his team to build Google's exclusive quantum chip.

In May 2016, Google released for the machine learning special research and development of the TPU (tensor processing unit) chip. The TPU chip is more robust in reduced computational accuracy and uses more sophisticated and high-power machine learning models. By quickly applying these models, users get more accurate results. Google claims that TPU will machine learning ability increased three generations, TPU Moore will advance seven years.

China IT News APP

Download China IT News APP

Please rate this news

The average score will be displayed after you score.

Post comment

Do not see clearly? Click for a new code.

User comments