Create Android for AI World _ Cut into artificial intelligence from different angles

In 2011, before Apple's founder Steve Jobs died, Larry Page, who was 38 years old, advised Google to do too much. Page replied: "You have not done enough. If we are just doing what we did before, and not trying new things, I will have a sense of guilt."

Jeff Dean, the head of Google Brain and a key figure in Google’s growth, confirmed this in an exclusive interview with Caijing reporters this year. He said that Google's exploration of the future is thorough and comprehensive. "Every brick we have to open up and look at."

When other companies are chasing hot spots, Google's goal is to eliminate blind spots.

However, Page admits that Jobs is right from a certain point of view: he can only manage so many businesses, and there are too many things that will cause problems in the process.

In order to support Google’s huge market with a market value of nearly $650 billion, it continues to run fast and avoid wrestling. In August 2015, Page announced the establishment of a holding company called “Alphabet”, which became the controlling parent of Google. Google" is only responsible for search business, Android, YouTube and several other businesses.

Parallel to Google, there are a series of wholly-owned small companies that are independent of their business lines. These companies are at the forefront of innovation and are highly innovative, most notably unmanned.

This restructuring has lasted for two years. In September, a document submitted by Google to US regulators showed that Alphabet created a holding company called XXVI Holdings, which included Alphabet and all its companies. In addition, Google itself will become a limited liability company, and Alphabet believes that this is more suitable for its status as a subsidiary.

Google officials said the move showed that their restructuring from 2015 was finally completed.

The goal of the reorganization is to reduce the burden and enhance the vitality of the behemoth. When Page announced the reorganization decision, he said that he should guard against Google becoming a well-behaved and secure company. “You need to make yourself feel restless in order to stay connected to the industry.”

Alphabet has a unique meaning. Alpha represents the performance advantage of a portfolio relative to the market index in a given time. Bet means bet, and the explanation is that the new company is established and organizationally adjusted, which is a proactive bet with a higher return than the broader market.

The chip in the valley singer is AI. In the past two years, Google has also adjusted its technical and business strategy direction, from the “Mobile First” strategy to “AI First”.

Jeff Dean told Caijing reporter that AI is not a business, nor a goal for Google. It is a reinvention of Google's ecological blood circulation system. Every product business of Google will be based on AI.

He believes that Google will continue to be a company that can maintain high growth in the future, depending on two things: First, Google has the most talents in the world; second, it can continue to create more than one billion people in the world. The product used.

When it comes to light, when a company dominates the market, it must start looking for new growth points. Google’s three-year restructuring is also based on this August 10, 2015, Google announced the reorganization into Alphabet and adopted a new operating structure; on October 2 of the same year, Google announced that from the fourth quarter of 2015, Alphabet’s earnings were divided. Google and other business parts.

The special feature of the Alphabet business structure is that the holding company was originally born on the basis of Google. Since the second quarter of 2015, the company has changed its architecture. Google has occupied the bulk of the overall revenue of Alphabet, and its proportion has remained at 99%. Above (except for the fourth quarter of 2016, this figure is 98.995%). Parallel with Google is the "crazy" innovative business.

Google’s business revenue is divided into three parts – Google’s own advertising revenue, Google’s advertising revenue, and other revenue. According to the financial report data for the second quarter of 2017, the corresponding share of the three is 71.52%, 16.49%, and 11.99%. In other words, advertising revenue accounts for nearly 88% of Google’s revenue.

This is why the outside world has been criticizing Google for being "actually an advertising company."

After the announcement of the restructuring, Google’s market capitalization has soared by $25 billion, and the overall financial situation of Alphabet has also grown steadily, especially in terms of revenue. Revenue in the second quarter of 2017 reached $26 billion, a 21% increase over the same period last year. This figure is better than Wall Street's optimistic estimate of $25.6 billion.

Some investors and analysts who have been tracking the performance of the Alphabet market for a long time said that Alphabet is already a huge super elephant, but it can still maintain double-digit revenue growth and keep running. It is terrible.

From a capital perspective, the benefits of setting up Alphabet are obvious. Previously, analysts used Google's traditional large-scale company valuation method to calculate valuation based on overall revenue, profit, growth and earnings per share. This approach is quite unfavorable for the current stage of Google investment layout innovation business. In the early days, these businesses not only failed to bring revenue to Google, but also brought huge costs and diluted profits.

Subsidiaries that are currently independent include Calico, a subsidiary that studies how to extend human life; Access & Energy , a high-speed fiber optic network ; Life Science, a smart health company ; Nest, a smart home company acquired in 2014 for $3.2 billion, Google Glass is also being The department; Google's core research department X Lab, after Google's many new business comes from this, including Wing drones, unmanned vehicles and Google glasses, in addition to Google venture capital business GV and Capital.

If these high-risk crazy projects always need to be carried out under the framework of Google, they will be restricted by investors and Wall Street, and they are extremely free.

With the new architecture, analysts can use another SOP (Sum Of Parts) valuation method that is commonly used in diversified holding companies. The advantage of this valuation is that the sum of the valuations of each subsidiary can be used. To calculate the overall value of the holding company, and guide investors to focus on innovative business.

But from a business perspective, Google says it is designed to allow other "crazy" businesses to have better room to grow.

Each enterprise under the control of the holding company is legally an independent economic entity. They are completely independent of the economic responsibility of the holding company and have no joint responsibility with each other. Therefore, the risk responsibilities of each company in the holding company will not be passed on to each other.

In addition, this loose business architecture is essentially an independent line of business, with clear responsibilities, making each company's operations more transparent, and the cost and business performance of each subsidiary at a glance. Encourage innovation and survive the fittest.

Li Kaifu, a former Google executive, publicly commented on the matter. He believes that Google’s restructuring is mainly to respond to Wall Street’s criticism that the company is not focused on opacity (investing in health, longevity, and automotive), providing operational data for each subsidiary, and the second is to reduce Outside of Google, the negative impact on the main business (such as losses caused by the automotive business, litigation, etc.), the third is the Google brand for Internet business, other aspects of the business should have other brands, and there are more opportunities for internal promotion.

The CEO of Alphabet is Page, and the chairman of the board is Sergey Brin. Their main responsibility is to provide as much help as possible to the following subsidiaries, plus the resources of the entire group and the performance evaluation of each subsidiary ExcuTIve team.

At this point, Google has at least loosened its organizational structure, and all explorations of the future of mankind have been moved to Alphabet, forming a parallel symbiosis with Google.

Create Android for AI World?

Whether it is to do a big 1% innovation business, or continue to maintain 99% of business growth, Google needs to find a new growth point, Google will put the next growth point on the AI

Jeff Dean emphasized to the Caijing reporter that AI will not be a business or a means, it is a capability.

He said that at Google, don't think of AI as something that an isolated R&D staff does. It runs through all Google departments and companies.

Google currently has the shadow of artificial intelligence in all its products.

The Google search business is a source of revenue for Google and even Alphabet. Over the years, Google has continued to use AI to reshape its search business. Third-party market research organization StatCounter

According to statistics from Global Stats in August this year, Google search accounted for 91.65% of the search market, and Microsoft Bing, which ranked second, accounted for only 2.5%.

According to data from third-party market research firm Smart Insights, the search volume on the mobile side is growing rapidly in recent years and has surpassed the desktop. Search volume from mobile devices also surpassed desktop browsers in 2015.

Closely related to the shift in mobile search volume to the mobile side is the ever-expanding mobile online advertising market which is often directly related to Google's money bag. Beginning in October last year, Google began to tilt the search business to the mobile side, and continuously upgrade the mobile search function and user experience through artificial intelligence technology.

There are many Google products that have been completely changed by machine learning/artificial intelligence, including YouTube, Maps, Android, Chrome, and more.

In the innovative business area where revenue is currently only 1% of Alphabet, AI is the genes of these businesses.

The AI ​​will become blood, running through every capillary in Google's body.

Google emphasizes AI productization, and more than a dozen Google products and services, including Google Translate, have been rewritten with AI algorithms. Google has also launched the "AI first" strategy, in addition to transforming the Google systemic circulation system and transforming the world with AI.

Regarding the "reform of the world", Google has done this in the era of mobile Internet: more than 85% of the Android operating system on the smartphones in the world has completely changed the way people use mobile terminals.

In the past two years, Google has put together the machine learning technology developed and adopted in the company, named TensorFlow. This is a framework that includes many common deep learning techniques, features, and examples that are used by almost all of Google's products.

A number of Google respondents, including Jeff Dean, told Caijing that Google wants to make TensorFlow an "Android" in the era of artificial intelligence. From a business perspective, this statement holds; from a technical perspective, TensorFlow is an open source learning framework, similar to Java, while Android is an open source operating system.

TensorFlow was formerly known as Google's first generation deep learning system, DistBelief. But DistBelief is just a Google internal system that is closely tied to Google's internal infrastructure and focuses solely on neural network algorithms. After some renovation, in November 2015, Google launched the open source deep learning platform TensorFlow (single version).

In February 2017, Google hosted the first TensorFlow Developer Summit in California and released the TensorFlow 1.0 release. The new version is five times faster than the first generation system in building and training neural networks. It can support multiple hardware platforms and even deploy on mobile phones.

TensorFlow is like Lego blocks that give AI capabilities. External developers and users can choose different building blocks according to different needs and build the AI ​​products they want.

At present, on the Github, the world's largest open source software community, Google TensorFlow is far ahead of other open source AI platforms by developers and code activity. Developers around the world upload more than 10,000 lines per week. A TensorFlow developer told Caijing that this is a manifestation of vitality for open source software.

Leiya, senior vice president of AI startups, told Finance. Compared with other open source deep learning platforms, TensorFlow has the advantages of strong community, complete functions, versatility and architecture design, easy deployment and suitable for large-scale products.

However, TensorFlow is also in its infancy. Deng Qi, a senior technician of Shanji Technology, told Caijing reporters that because of the pursuit of versatility, TensorFlow is not as good as other platforms in terms of flexibility and speed, interface design and other performance indicators. The platform leaves development opportunities.

Shanshu Technology is an AI company. After the previous review, we finally turned to the machine learning and numerical optimization platform based on the Python language. Strictly speaking, it does not belong to the TensorFlow camp.

Leya said that Google is very ambitious on TensorFlow. Open source TensorFlow is not just about posting source code, but to attract and lead developers around the world to update TensorFlow to become the deepest learning. The platform has become the de facto standard for the global AI development platform.

If you can do this, AI products around the world will continue to shed Google's blood and drain potential wealth to Google, just as most of today's smartphones carry the Google Android operating system.

Jeff Dean said that although artificial intelligence can't directly generate revenue for Google, other advertising revenues from Google's business using artificial intelligence technology can in turn subsidize R&D investment.

Google has therefore invested heavily in the TensorFlow community and has a strong publicity. In contrast, some open source deep learning platforms from academic institutions or individuals may be better than TensorFlow in some places, but it is difficult to guarantee continuous human and material resources in the future.

Many industry professionals told the Caijing reporter that Google's open source AI platform is also a wise move. The AI ​​industry is still in the early stage of development. It is difficult for Google to drive the industry solely because of its own power. As a platform manufacturer, it is imperative to make a big ecology.

Google open source TensorFlow has also prompted other big companies to open up their own deep learning platform. The entry barrier for the AI ​​industry has been greatly reduced, because anyone can take its source code for research and use, providing a large number of innovative AI applications. For example, local AI startups go out and ask TensorFlow to train voice assistants.

Michael A. Cusumano, a professor at the MIT Sloan School of Management, previously commented that even though the TensorFlow open source platform itself is not making money, if it succeeds, it will become Google's money-making machine.

“In the future, all walks of life need machine learning and artificial intelligence to help them develop their businesses, but there are very few companies in the world that really have the ability to hire machine learning scientists, and there may be only a few thousand.” Jeff Dean told the Caijing reporter. "The meaning of Google is also here."

Leja explained that if you want to use the more advanced features of TensorFlow, it is best to use Google's public cloud (not free). AI needs strong computing power, and Google Cloud has developed a AI-specific chip TPU to provide more powerful computing power. Google actually provides AI application developers with a full set of tools and services from chip to cloud to deep learning platform to peripheral software.

Of course, Google Cloud does not currently enter China. It is impossible for Chinese AI companies to use Google Cloud across borders because data may be transmitted overseas and may violate the newly released Cyber ​​Security Law. This left a buffer zone for Baidu, Tencent and Ali to develop their own AI platform and cloud services.

Google has always wanted to make TensorFlow an Android system in the AI ​​era.

Going out to ask is an AI startup company that provides voice assistant products and services. Its founder and CEO Li Zhifei told the Caijing reporter that Google Andrew’s successful experience has three points: First, a standard is defined, and everyone develops according to this standard. Application; Second, there are active communities, developers in the community to answer questions, share development experience, and even do not need Google's official help; Third, there is a rich upstream and downstream industry chain support, upstream there are a variety of mobile phones With a large number of application developers downstream, and developers can make money.

TensorFlow is also going in this direction. Compared with Android, TensorFlow is trying to become a de facto standard, and the degree of community construction is also ok, but the upstream and downstream industry chains are not rich. More importantly, other technology giants and research institutions are also striving for opportunities in this area, such as Microsoft's DMTK, Facebook's Torchnet, Baidu's Paddle, and so on.

Recently, Microsoft Group and Facebook announced a strategic partnership, the two sides will jointly establish an open neural network communication (ONNX) format, AI applications will be able to switch freely on both open source platforms. Many industry insiders said that this is the embodiment of the two companies joining forces to prevent Google from becoming a big one.

Li Zhifei believes that in the short term, all open source AI platforms have opportunities, but in the long run, there will be no more than two or three platform-level products. In addition to Google, Amazon and Microsoft also have opportunities, because both have powerful public clouds, while AI is away. Do not open powerful computing power. Leya believes that the winner must be a deep learning platform that is easy to learn and use.

As for the Chinese AI company, Li Zhifei believes that a large number of AI applications will be born in the future, but it is difficult to produce a world-class deep learning platform vendor that can compete with Google in the near future.

Google's goal is not only to turn AI into the genes of all products in the Google system, but the greater ambition is to shape a new "general artificial intelligence" order, and Google is the leader of this order. Jeff Dean predicts that the creation of a universal artificial intelligence world can take up to 15 years.

Innovation path under the new architecture

For more than a decade, Google has been good at innovative products for individual consumers, but to penetrate the industry and develop industry users, Google needs more changes. Lily Peng is a Google product manager. She looks very young, but she is a twin. mother. She is also a medical scientist, and her team's internal mission in Google is to combine deep learning techniques with medicine.

There are a lot of cross-border talents like Lily inside Google. In order to ensure that technology can be deployed in different sub-sectors, talents with technical skills and industry experience are essential.

Lily's team started working with hospitals in India. In India, many hospitals have found that many people with diabetes cause retinopathy, and because of the lack of professionally trained ophthalmologists, it is not possible to efficiently interpret the patient's fundus image.

Lily's team trained algorithms for some doctors in India, including image recognition technology to help doctors interpret the patient's fundus image. "Our algorithm results are almost the same as those obtained by professional ophthalmologists," she told the Caijing reporter.

The technology is based on the Tensorflow platform, which combines the TensorFlow platform algorithms to help not only perform fundus examinations, but also image inspection of cancer.

As a female employee, Lily can feel the company's support for her. “Google gave me six months of maternity leave,” Lily told the Caijing reporter. “After a full rest, I quickly joined the team.”

Prior to joining Google, Lily worked at a startup. At Google, she doesn't need to consider the issue of making money like an ordinary product manager. “We only need to be a researcher and concentrate on developing medical solutions.”

Google has a tradition that has been imitated by the outside world – in order to motivate employees to innovate, each employee has 20% free time to focus on any area of ​​interest. In 2004, Page and Sergey Brin emphasized this famous management philosophy in Google’s IPO prospectus. “We encourage employees to use 20% of their time to do whatever they think about Google. Good things."

This really brings value to Google's innovation. Two engineers from Google's Paris department in France used Google's "20% time" to use cardboard as the main material and spent six months developing a VR head display. In June 2014, Google launched this head-mounted display, named "Cardboard", and this head display has become the originator of many mobile VR companies competing to imitate.

Even in the VR industry, this has become a major category of VR headsets - low-cost mobile VR boxes are collectively called Cardboard.

One fact that needs to be clarified is that not all Google employees agree with the "20% time" rule.

Yahoo CEO Marissa Mayer used to be an executive at Google. After she joined Yahoo, many employees asked if she would also implement such a policy at Yahoo. “I feel so funny,” Marissa Mayer said at the Yahoo staff conference. “That 20% of the time means you will spend 120% of your time at work.”

In 2015, according to Google's internal statistics, only 10% of employees use this 20% of the time, and the company does not allow employees' own work to be affected before the project being studied is truly fruitful.

When there are no good ideas, more people choose to do their best to complete 100% of their jobs.

When Lily was asked by Caijing reporter how to use her 20% time, her answer was: "I am 100% enthusiastic about my current job. I will not spend 20% more time doing other things. Things."

This has ensured that Google employees' core energy is focused on their work. The “20%” innovation capability that has been utilized is carefully thought out and selected to ensure the precision and success rate of innovation.

A Google China employee told the Caijing reporter that after the Alphabet structure was built, the goals of each subsidiary were more clear, and the responsibilities and goals were more clearly communicated to individuals. For individuals who are interested in developing “20%” time, This is a good thing.

Laszlo Bock, former senior vice president of Google’s human operations, is also the author of the book “Redefining the team: how Google works.” He wrote in this book: “In a way, 20% of the time is better than its Reality is more important, it can play a role outside of normal corporate management, especially for those who are most talented and creative."

People think that Apple is used to look up, and Google can copy it. The successful combination of Apple products and art has allowed Apple and Jobs to stand at the commanding heights of many technology companies. But Google is different. Google's management system, corporate culture, and corporate strategy are not secrets. There is no Steve's genius and extreme artistic talent.

There are indeed many companies that are imitating Google. In April last year, Baidu also announced the reorganization. Baidu Chairman and CEO Li Yanhong announced through internal mail that Baidu established "Baidu Search Company", which consists of Search Business Group (SSG), Mobile Service Group (MSG) and Demi Business Unit. In February 2015, Baidu was tuned to Mobile Service Business Group (MSG), Search Business Group (SSG) and Emerging Business Group (EBG). In December 2015, Baidu announced the formation of the Financial Services Business Group (FSG).

This restructuring model is similar to Google's, separating the mature business from the innovative business, focusing on the core business while ensuring innovation.

But Jeff Dean emphasized to the Caijing reporter that Google’s exploration is comprehensive. A business structure that keeps pace with the times needs to be aligned with clear and clear strategic goals. But in today's world, "Most of the companies have chosen to follow the hot spot, which allows the company to achieve rapid business growth in a short period of time, but can not really lead the trend."

If a new architecture is not established, a scientist company like DeepMind may only be a "rocket fuel" player, but now DeepMind is going to start going out. "If Google is a rocket, DeepMind is fuel." DeepMind founder Jamies Ha Sabis told the Caijing reporter.

In the next stage, Google is based on AI, and the subsidiary company DeepMind is of great significance to Google's AI strategy.

The company started out as an artificial intelligence laboratory in London, England, with a research direction developing universal self-learning algorithms. In 2011, Hasabis established DeepMind under the investment of Elon Musk and others. In 2014, Google acquired the company for £400 million (approximately $650 million), when the company had only 50 employees.

Google artificial intelligence Go application "AlphaGo" defeated Korean player Li Shishi in 2016, thanks to DeepMind. However, few people know the current position of this team within Google, especially after AlphaGo, where will this scientist company go?

To answer this question, you must first understand why one of the world's most valuable companies, DeepMind chose Google among multiple olive branches.

The founder of the World Wide Web, Bonus Lee, called Hasabis "the smartest person on earth." He was a child prodigy when he was a child, and he began to think about the combination of neural networks and computers in his teens. And Hasabis thinks that he is "born for artificial intelligence."

However, operating companies are different from academic research in universities, and ensuring that companies have sufficient funds to develop is the primary focus. Even the genius Hasabis felt that he was weak, and three years later he chose to sell the company to Google.

At that time, in addition to Google, Facebook also had the willingness to acquire the company, but Kasabis believes that Page has 120 points of enthusiasm for artificial intelligence, and Zuckerberg lacks the basics, and communication with people is the ultimate goal. . Hasabis believes that he is more scented with Page.

Google’s advertising business has brought huge revenues to the company, and research teams like DeepMind have plenty of freedom to grow.

When Google acquired DeepMind, it has more than 10 years of experience in computer infrastructure and engineering infrastructure. In early 2016, Kasabis mentioned in an interview with the US media that they could simultaneously parallel 1 million experiments due to Google’s support. The research content has been greatly increased, and the research speed has been greatly improved.

With DeepMind, Google quickly enhanced its deep learning algorithm capabilities, which is the logic of Google's acquisition of technology companies. But the two companies have different genes. Kasabis told Caijing that "Google is an engineer-oriented model and DeepMind is scientist-oriented." He emphasized that DeepMind's team of scientists will be in the future. Positioning will not change.

At the same time, DeepMind quickly became the aortic blood vessel of Google's new ecosystem. DeepMind has begun to use Google's own AI research and development results to solve practical problems for Google.

In July 2016, DeepMind announced that DeepMind has successfully helped Google's data center cooling system save 40% of its electricity using deep learning algorithms. The cooling system is an essential accessory for large industrial equipment, including chillers, pumps, cooling towers, etc., to maintain the normal operation of the data center server.

“There is too much electricity on Google’s data center equipment, and a few percentage points means millions of dollars a year,” said Hasabis. “The algorithm for saving electricity saves the company a lot of money every minute.”

The algorithm's development team only has five or six people, including Mustafa Suleyman, R&D engineer Rich Evans, and data center engineer Jim Gao, which took only two or three months.

Some of DeepMind's R&D projects have been used for energy infrastructure management, medical systems, and improvements to clean water sources, and it has begun to generate revenue.

Hassabis told the Caijing reporter that in the next stage, 75% of DeepMind's entire team will focus on artificial intelligence technology development, and another 25% focus on applications.

Another co-founder of DeepMind, Mustafa, dominated the commercialization of the company's technology, which he believes is a balance. "We have a lot of new technologies in the laboratory, which are valuable resources. The commercialization of these technologies can not only benefit the society, but a company must also have a sustainable source of commercial profits." He told Caijing.

According to him, DeepMind is currently composed of two teams, Kasabis is responsible for the R&D team, and he is responsible for the technology application business. The technical application team is subdivided into three groups, one serving Google’s industry and Google’s other business units; the second is the medical group, which has partnered with the UK’s National Health Service; the third is the Energy Group, currently Still in the startup phase.

“This grouping is to facilitate better cooperation with experts in different fields,” he said.

DeepMind has also begun to enter the medical market and profit from it. In November 2015, it signed a five-year contract with the Royal Free Hospital of London, UK, to handle the medical records of 1.7 million patients.

This is in line with the purpose of Google's transformation. After the parent company Alphabet was founded, the entire company's balance sheet began to be divided into accounts, the business was independent, clear responsibilities and achievements, and improve innovation efficiency.

A capital market analyst analyzed and evaluated the "Financial" reporter. If it has been in the previous structure, DeepMind can hardly measure the value of Google's technology blood transfusion, and it is difficult for DeepMind to have a clear long-term research goal that is consistent with Google's strategy. .

One of the more important reasons is that senior scientists and strong computing power are important to the development of AI business, but it is equally important for DeepMind to go out and get real-world data. These data that go deep into the industry, such as medical care, are just missing data from Google. source. As a search and social gene Internet company, the huge amount of data in the valley singer is mostly open social attribute data.

Alphabet has a number of innovative subsidiaries, some of which are known for their technology, and some with business innovation, but in the next phase, their goals are the same - sinking into the industry, continue to mine data gold, analysis, Use and generate value.

Another senior person from China's top global artificial intelligence company told Caijing that the current amount of data flowing in an open environment such as a social platform accounted for only 20% of the total data, and 80% of the data was hidden in various industries. Within them, they are "gold glittering and have to be excavated."

Hassabis has repeatedly stressed to the Caijing reporter that, to a certain extent, DeepMind operates independently of Google, "just like two parallel companies."

The next golden age?

Google still hasn't stopped today, its golden age is not over, it still brings huge imagination.

Google was listed on the Nasdaq in 2004. At the time, the company had just been established for six years, and its valuation was soaring, reaching $23 billion and a price-to-earnings ratio of 80.

However, Google’s listing achieved $2.7 billion in revenue in the year, with a profit of $286 million. After 13 years, the market value has reached 64.32 billion US dollars, revenue of 90.3 billion US dollars, profits of 19.5 billion US dollars, and the market value has increased by 27 times.

In the global market, only the technology companies that have entered the "$500 billion senior club" are Apple, Google's parent company Alphabet and Microsoft, and the technology company Facebook is also expected to join the ranks in the near future.

For more than ten years, Google has developed at a fast pace and has a large scale, which is unexpected from the outside world. In 2011, Wall Street venture capitalists warned that Google’s “golden age” may have become history. In other words, Google is likely to take a “downhill”.

At that time, Google had been listed for seven years, with revenues exceeding $37.9 billion and annual growth of over 29%. Both the total and the growth rate, the outside world believes that the company's business development has reached its peak.

The skeptics believe that Google’s revenue at the time was too dependent on search advertising, and that the emerging social networking sites were also likely to divert Google’s traffic and advertisers. But it didn't happen. In 2012, Google's total revenue actually exceeded $50.2 billion, an increase of 32% year-on-year.

In the following years, Google’s revenue and profits have maintained rapid growth. By 2016, Google’s revenue reached $90.3 billion and its profit was $19.5 billion.

At the moment, Google is not the only company that values ​​AI. Other technology giants and some traditional companies have been investing in robotics and AI functions. Such as Amazon, Facebook, Microsoft, IBM and OpenAI.

In the 2016 annual report submitted to the US Securities and Exchange Commission in February this year, Google's parent company, Alphabet, mentioned that its current major competitors include Apple, Netflix and Hulu, while highlighting new businesses such as consumer hardware manufacturing and enterprise. Competitive risks faced by cloud computing and digital assistants.

In Google’s most wanted cloud service area, Google’s strong competitors are Amazon and Microsoft. Under the leadership of executive Diane Greene, Google’s cloud computing business is hoping to become a true service for large enterprise consumers. Last year, Alphabet also mentioned the cloud business in the annual report, but focusing on the company is still a new formulation.

According to data from third-party research firm Canalys, Google’s cloud computing revenues were around $700 million in the first quarter of this year, a big difference from the leader’s $3.6 billion. However, in the second quarter, the number of Google Cloud Service contracts (orders above $500,000) was three times that of the same period last year. In the same period, “non-advertising revenue”, which included revenue from cloud services, increased by 50% to reach US$3 billion (advertising revenue of US$22.7 billion), indicating that Google’s cloud services business is growing rapidly.

However, for the current Google, cloud computing revenue is too small, and the profit margin is very low, it will not have much impact on Google's overall revenue and stock price.但谷歌云此前一直为自身业务所用，“AI First”战略之后，提出了以“TensorFlow、TPU和云”三剑客去争抢外部的目标，这就不得不令亚马逊AWS和微软Azure侧目了。

在谷歌的核心现金流业务之一，Youtube的竞争对手是Facebook、Netflix、Amazon和Hulu。今年6月底，Youtube宣布活跃用户数已经超过15亿，这个数字已经十分可怕，相对于中国的微信接近9亿的用户总数优势十分明显。但它最强劲的竞争对手Facebook活跃用户已经超过20亿。

社交网站的竞争核心已经从用户规模转移到了付费视频，Youtube今年来已经为此在技术和付费订阅模式上做了不少改进。

不过，付费业务是所有视频社交网站需要跨越的一个山头，谷歌管理层目前对这个业务抱“谨慎的乐观态度”。

Alphabet架构重整之后，除Google之外的其他业务收入均不会单独列出，不过，根据华尔街投行分析，2016年Youtube的收入大约在100亿美元左右，而且增长速度很快。他们预估Youtube的营收未来几年内很有可能达到300亿美元以上，它将成为谷歌营收达到1500亿美元的关键因素。

在数字助手服务领域，谷歌设定的竞争对手是苹果、亚马逊、Facebook和微软。

去年，谷歌发布了数字助理Google Assistant，这是一个基于语音的信息服务。谷歌希望能将Google Assistant嵌入到尽可能多的设备上。这与苹果的Siri、亚马逊的Alexa和微软的Cortana构成了直接的竞争关系。Facebook也有类似的服务Messenger。

在设计、制造和销售消费电子产品领域，谷歌定位的竞争对手包括了苹果、亚马逊在内的多个巨头公司。

硬件从来都是谷歌的短板，尽管从2016年来，谷歌已经把制造一系列新硬件产品提上日程，不过，AI的本质是服务，硬件只是服务的载体，最好的人工智能服务会变得成为“隐形”的服务、云端的服务，而硬件也会微缩化，从这点上看，谷歌在这个领域暂时不会有太大的突破。

需要强调的是，谷歌以上领域要继续增长，均以AI为核心动力。谷歌在AI领域的技术积累优势已经无需质疑，但能否将AI在谷歌这个大盘子内最高效盘活并尽快产生良性的全身血液循环，是谷歌能否继续高歌猛进的关键因素。

战略既定，诸多事项有待发力。多位接受《财经》记者采访的谷歌高层人士均表示，和其他公司竞争更多人才是谷歌在下一轮商业和技术竞争中的一个核心指标，尤其在其他条件基本相似的情况下。

Jeff Dean对《财经》记者说，谷歌目前对所需人才的定义有三种，一类是那些能真正理解并运用人工智能工具的技术工程师；其次，是各行各业的专业跨界人才，如精通医学和技术的技术专家；最重要的是，有创新意识和商业头脑的人才，能有效地将人工智能技术更广泛地应用到各个行业的人才。

这和谷歌在过去十数年专注引进计算机高精尖人才的路数已经完全不同，也与谷歌下一阶段希望以行业为突破口，进入企业级市场的全新目标契合。

不过，这也是亚马逊、苹果同样看到的机会。谁能率先建立起一套成规模气候的人才机制，谁就有可能率先突破这个领域。

谷歌过去的故事和今天的进击告诉我们，进攻是最好的防守。谷歌成立之初以搜索引擎挑战者入局，时至今日，当初的被挑战者雅虎早已淡出市场，几乎与它同时成立的百度虽然也布局了AI技术，希望在下一阶段成为一个“AI All In”的公司，但时至今日仍然是一个以搜索引擎为主要面目出现的公司，两家公司市值相差将近8倍，AI技术上的布局更不在一个量级。

谷歌和百度在这十余年间的浮沉进退折射了一家科技公司在多次技术迭代浪潮中的路径选择重要性，以及一套“生于忧患死于安乐”的商业立世准则。但谷歌的未来之路也并非坦途，毕竟其它选手也在以不同的角度切入未来人工智能世界。

Onlyrelx mini500

Disposable vape pen Onlyrelx 500puffs is portable and fashion disposable electronic cigarette, disposable ecigs pen are trending featured vape pen for vapors as it's safety and easy to use. Disposable vape pod,disposable vape, wholesale vape,vape wholesale,vape pen manufacturer and supplier.disposable vape pen,disposable electronic cigarette,disposable ecigs pen,disposable ecigs stick,disposable e-cigs pen,disposable vape factory,disposable vape pod,disposable vape device,vape pen,vape stick, vape wholesale,wholesale vape,customized dispsoable vape pen,customized vape pen,OEM&ODM disposable ecigs pen,disposable electronic cigarette wholesale, wholesale disposable electronic cigarette,distribute vape pen,vape pen distribute,high quality vape pen,high quality vape pod,rechargeable disposable vape pen,refillable vape pen,refilling electronic cigarette,rechargeable disposable electronic cigarette,refillable vape pod,disposable refillable ecigs,disposable refilling e-cigs pen,refillable e-cigs pen

Onlyrelx mini500,Onlyrelx vape bar,Mini Vaporizer Onlyrelx,Original Onlyrelx Vape Pen,disposable vape pen

Shenzhen Onlyrelx Technology Co.,Ltd , https://www.onlyrelxtech.com