News classification
Contact us
- Add: No. 9, North Fourth Ring Road, Haidian District, Beijing. It mainly includes face recognition, living detection, ID card recognition, bank card recognition, business card recognition, license plate recognition, OCR recognition, and intelligent recognition technology.
- Tel: 13146317170 廖经理
- Fax:
- Email: 398017534@qq.com
Artificial intelligence bottom platform
Artificial intelligence bottom platform
The time of machine learning and artificial intelligence had come. The development of big data, large capacity storage, elastic computing and all kinds of algorithms, especially in the field of deep learning, brings various innovative applications of brain holes.
In the complex strategic games such as weiqi, machines have been better than humans. Application of image recognition, speech recognition and so on is be nothing difficult. The voice intelligent assistant has begun to advance, fully automatic driving test on the road. But many of the discussions about machine learning / artificial intelligence in the near future are algorithms and applications, and few discussions have touched the underlying infrastructure.
In the early stages of computing, only the compilation of speech experts, compiler experts and operating system experts can be used to develop simple applications. The current situation is also very similar. Only if you get a Ph.D. in statistics or distributed system, can you know how to develop AI system and deploy it in a large scale. The missing link is to speed up the general tools of artificial intelligence development. Therefore, only the most elite engineering team will have the perfect ability to do this work.
On the other hand, as for the innovation of machine learning technology, the development of basic equipment is also very backward. In simple terms, systems and tools that are used as the basis for current machine learning are not suitable for the evolution of future intelligent applications. In the future, the industry needs new tools to release the potential of artificial intelligence, making artificial intelligence more approachable and more applicable. So in the category of basic equipment entrepreneurship, the modules needed for the development of intelligent systems will be a big gold mine in the future.
From basic equipment 1 to basic equipment 2
The relationship between the application and the basic equipment is very ingenious, limiting each other and advancing each other.
The expansion of hardware and system software brings a new type of application. These applications are perfected and mature from time to time, thus making higher requests for the underlying resources and forcing the underlying infrastructure to innovate. In turn, the optimization, innovation, and cost performance of the basic equipment will also bring the overthrow of the application, providing unprecedented experience to the users. A typical example is from slides to PPT, to a variety of online photo social platforms, such as Pinterest.
At the beginning of this century, the commercial Internet was launched based on the x86 instruction set from Intel, the standardized operation system from Microsoft, the relational database from Oracle, the Ethernet device from CISCO, and the network storage tool from EMC. Amazon, eBay, YAHOO, and the original version of Google and Facebook are based on these basic devices. This is the "basic equipment 1" in the science and technology industry.
But with the gradual maturity of the Internet, the total number of Internet users has increased from 16 million in 1995 to about 3000000000 at the end of 2015, and the application's request for scope and performance has also been greatly improved. The technology of the "client / server" period is no longer suitable for the needs of the Internet giants, whether in terms of feasibility or cost performance.
As a result, the Internet Co began to be self-sufficient. Google, Facebook, and Amazon defined a new type of infrastructure by relying on its own technical expertise and the academic pause. Such a basic device has the following features: extensible, programmable, usually open source, and low cost. Related technologies, including Linux, KVM, Xen, Docker, Kubernetes, Mesos, MySQL, MongoDB, Kafka, Hadoop, and Spark, define the period of cloud computing. This is also known as the "infrastructure 2" of the technology industry.
The most central technology of this generation is designed to enable the Internet to mask billions of end users and acquire and store information from so many users in an efficient way. Therefore, the innovation of "basic equipment 2" has led to a substantial increase in the amount of data. With the deployment of concurrent parallel computing technology and algorithms, we have seen the development of current machine learning.
Basic equipment 3: towards an intelligent system
The ultimate question of the "basic equipment 2" period is, "how do we connect the world?" And the current problem is more: "how do we understand the world?"
这其中的差别,即“衔接”和“认知”,可以解释人工智能与上代软件的关键不同。 The code's own "cognitive talent" overturns traditional programming. In the traditional application, the program logic is written to the dead, and in the artificial intelligence application, the algorithm draws the logic through the analysis of the large data. Subsequently, these logic are used for decision-making and prediction.
The result of this is the "intelligent" application. But in theory, this kind of application needs a lot of data and a huge amount of computing resources. These restrictions make it difficult to general-purpose the artificial intelligence, which corresponds to the computational paradigm proposed by von Neumann 70 years ago. So, AI represents a basic new architecture that entreat us to rethink the infrastructure, tools, and development theories.
So far, the research and innovation in the field of artificial intelligence still focus on new algorithms, model training techniques and optimization methods. In addition, only a small part of the code in AI system is used for learning and prediction, and the most convenient part is to prepare data and develop functions, so as to enable distributed basic devices to function and execute tasks in a wide range.
If you want to develop and deploy AI applications successfully, you need a harmonious multiple discrete system to design a precise process. First of all, you need to digest data, get data from the tag. Then, in order to complete the prediction, you must be sure of the proper features. Finally, developers need to exercise the model and
In the complex strategic games such as weiqi, machines have been better than humans. Application of image recognition, speech recognition and so on is be nothing difficult. The voice intelligent assistant has begun to advance, fully automatic driving test on the road. But many of the discussions about machine learning / artificial intelligence in the near future are algorithms and applications, and few discussions have touched the underlying infrastructure.
In the early stages of computing, only the compilation of speech experts, compiler experts and operating system experts can be used to develop simple applications. The current situation is also very similar. Only if you get a Ph.D. in statistics or distributed system, can you know how to develop AI system and deploy it in a large scale. The missing link is to speed up the general tools of artificial intelligence development. Therefore, only the most elite engineering team will have the perfect ability to do this work.
On the other hand, as for the innovation of machine learning technology, the development of basic equipment is also very backward. In simple terms, systems and tools that are used as the basis for current machine learning are not suitable for the evolution of future intelligent applications. In the future, the industry needs new tools to release the potential of artificial intelligence, making artificial intelligence more approachable and more applicable. So in the category of basic equipment entrepreneurship, the modules needed for the development of intelligent systems will be a big gold mine in the future.
From basic equipment 1 to basic equipment 2
The relationship between the application and the basic equipment is very ingenious, limiting each other and advancing each other.
The expansion of hardware and system software brings a new type of application. These applications are perfected and mature from time to time, thus making higher requests for the underlying resources and forcing the underlying infrastructure to innovate. In turn, the optimization, innovation, and cost performance of the basic equipment will also bring the overthrow of the application, providing unprecedented experience to the users. A typical example is from slides to PPT, to a variety of online photo social platforms, such as Pinterest.
At the beginning of this century, the commercial Internet was launched based on the x86 instruction set from Intel, the standardized operation system from Microsoft, the relational database from Oracle, the Ethernet device from CISCO, and the network storage tool from EMC. Amazon, eBay, YAHOO, and the original version of Google and Facebook are based on these basic devices. This is the "basic equipment 1" in the science and technology industry.
But with the gradual maturity of the Internet, the total number of Internet users has increased from 16 million in 1995 to about 3000000000 at the end of 2015, and the application's request for scope and performance has also been greatly improved. The technology of the "client / server" period is no longer suitable for the needs of the Internet giants, whether in terms of feasibility or cost performance.
As a result, the Internet Co began to be self-sufficient. Google, Facebook, and Amazon defined a new type of infrastructure by relying on its own technical expertise and the academic pause. Such a basic device has the following features: extensible, programmable, usually open source, and low cost. Related technologies, including Linux, KVM, Xen, Docker, Kubernetes, Mesos, MySQL, MongoDB, Kafka, Hadoop, and Spark, define the period of cloud computing. This is also known as the "infrastructure 2" of the technology industry.
The most central technology of this generation is designed to enable the Internet to mask billions of end users and acquire and store information from so many users in an efficient way. Therefore, the innovation of "basic equipment 2" has led to a substantial increase in the amount of data. With the deployment of concurrent parallel computing technology and algorithms, we have seen the development of current machine learning.
Basic equipment 3: towards an intelligent system
The ultimate question of the "basic equipment 2" period is, "how do we connect the world?" And the current problem is more: "how do we understand the world?"
这其中的差别,即“衔接”和“认知”,可以解释人工智能与上代软件的关键不同。 The code's own "cognitive talent" overturns traditional programming. In the traditional application, the program logic is written to the dead, and in the artificial intelligence application, the algorithm draws the logic through the analysis of the large data. Subsequently, these logic are used for decision-making and prediction.
The result of this is the "intelligent" application. But in theory, this kind of application needs a lot of data and a huge amount of computing resources. These restrictions make it difficult to general-purpose the artificial intelligence, which corresponds to the computational paradigm proposed by von Neumann 70 years ago. So, AI represents a basic new architecture that entreat us to rethink the infrastructure, tools, and development theories.
So far, the research and innovation in the field of artificial intelligence still focus on new algorithms, model training techniques and optimization methods. In addition, only a small part of the code in AI system is used for learning and prediction, and the most convenient part is to prepare data and develop functions, so as to enable distributed basic devices to function and execute tasks in a wide range.
If you want to develop and deploy AI applications successfully, you need a harmonious multiple discrete system to design a precise process. First of all, you need to digest data, get data from the tag. Then, in order to complete the prediction, you must be sure of the proper features. Finally, developers need to exercise the model and