Is the so-called operating system who have the world, can be dominated by the application, under the control of the hardware, more importantly, the operating system is the core of information and knowledge control points, which is a world-class business opportunities. Microsoft is relying on the PC operating system monopoly, has become the world''s highest market value of technology companies, it is lost control of the operating system and completely lost the opportunity in the phone.
Today, apple and Google instead, in the mobile phone operating system occupies a leading position. Research firm Strategy Analytics released a report last November, the third quarter of last year, Apple Corp grabbed a global mobile phone profit of 91%, a record high. BMO Capital Markets is expected in the third quarter of last year, apple grabbed the global mobile phone industry''s profit of $103.6%, while Apple''s mobile phone market share of only 12%. Why go beyond 100% because there are a large number of manufacturers in the loss.
Where is China''s "operating system" in the artificial age
(photo: Strategy Analytics)
The past history shows that the operating system can get Tulong Dao, like Microsoft, apple that draw most of the profits of the industry chain, the company can be in an invincible position, the other may be the history of supporting and travelers.
So what is the new "operating system" in the artificial age?
To answer this question, first from a computational perspective, from the current development situation of artificial industry''s point of view, because the artificial algorithm involves a large number of matrix calculation and parallel numerical calculation, calculation of the next generation has been shown to migrate from the serial parallel computing trend. In the past, the computation is represented by CPU, which is optimized for serial instructions, while the future computation may be represented by GPU, which can be used for large-scale parallel computing.
The software determines hardware rules appear repeatedly in history, if large-scale parallel computing has become the mainstream, then dominate the computing machine learning framework may become a calculation of the operating system, a machine learning framework behind is a complete set of tools for developers and a large developer community, interaction the direct and application layer or other intermediate layer, the following is interact with a computing device.
"New York Times" quoted saying, "what is at stake is not artificial era innovation of these 00 pieces, but may be a new computing platform control."
TensorFlow as a representative of the distributed machine learning framework is a good example, Google is clearly in the open source of its depth learning algorithm to promote the TensorFlow industry as the de facto standard. TensorFlow is the core of the surrounding libraries and frameworks such as TensorFlow Serving, TF-Slim and other key components of its ecosystem. On application and algorithm, to keep down the hardware, vertical integration of pressure and absolute dominance, make the downstream companies lose strategy differences, and more and more dependent on the core components of the provider, to become a vassal.
Where is China''s "operating system" in the artificial age
TensorFlow (Google) is a distributed machine learning framework
Behind the machine learning framework is an industrial chain, but when the industry chain is a monopoly, it will erode the entire industry chain most of the profits.
In theory, a distributed machine learning framework can be used to determine the fate of hardware vendors. To seriously say, if one day TenslorFlow obtain a monopoly in the industry decided to gradually abandon the GPU support, or to optimization tend to support TPU, so the NVIDIA GPU manufacturers will face a huge threat.
Which is more likely to happen is if Google really TeensorFlow dominated the distributed machine learning framework, then Google hardware manufacturers certainly do not want a single large, and hope that they become no different hardware provider. The chip maker NVIDIA as an example, they will face two choices, either Google and the depth of cooperation for their work, either give up the direction of other chips.
Similarly, if the TensorFlow priority or depth optimization of some Google Cloud machine learning cloud computing related API, Google in this machine learning cloud areas will also make great advantages. In fact, the Google version of the internal TensorFlow and Google Cloud depth binding optimization, other vendors with open source TensorFlow in the performance of a disadvantage. Then to the application layer, the monopoly of the TensorFlow also have the right to decide to optimize some of the characteristics, so that the Google at the application level to gain advantage.
To further emphasize that individuals, companies and even the country''s independent development in the field of artificial may be controlled by the monopoly of the machine learning framework.
This concern is not alarmist, founder of &CEO, horizon robotics Baidu Deep Learning Institute (IDL), the founder of deep learning expert Yu Kai also expressed the same concerns about Silicon Valley spy, "let TensorFlow become artificial development platform occupies a dominant position in the world is dangerous to the world. Although this platform is currently open, but with the passage of time, labor become more and more powerful, the system will become extremely complex to lose transparency, and will become the world''s terrible data, computing, hardware, etc. the standard compiler makers. This will lead to an unhealthy ecology, hinder young people to master the freedom of technology, so that individuals, companies and even the country''s independent development in the field of artificial, eventually