Home > News content

Jia Yangqing officially takes the lead in Ali Open Source! How did the AI Architecture God come into being?

via:博客园     time:2019/9/27 13:45:21     readed:90

[Introduction to Xin Zhiyuan]Yesterday, Jia Yangqing released the exciting Alibaba Flying AI platform. Just now, Alibaba officially announced that Jia Yangqing will be the head of Alibaba Open Source Technical Committee. From the core author of the world-class framework to the head of Ali Open Source, what changes will this self-contained AI God bring to the open source world?

Just now, Jia Yangqing officially becameHead of Alibaba Open Source Technical CommitteeOpen source upgrading has become one of Ali's technology strategies.


Today, Jia Yangqing, who is on the stage, has a new identity: the head of Alibaba Open Source Technical Committee.

Open source this matter, led by Jia Yangqing, is of great significance. Because he is one of the three most popular open source in-depth learning frameworks Caffe,TensorFlow and PyTorch. Jia Yangqing introduced what Ali has done in open source. Ali is starting fromContribute to open source.

  • One is the deep contribution of existing projects.
  • Second, many of Ali's current autonomous projects will gradually be put into the open source community.


Ali not only actively embraces open source, but also actively nurtures open source. He has many core contributors in the fields of Linux, MySQL, JVM and other international open source projects.

In recent years, Ali has taken the lead in investing in independent and open source, continuously exporting self-research tools and products. At present, there are more than 1,000 open source projects, with more than 660,000 benchmarks, and more than 10,000 technicians in Ali contributing code to these projects.

These projects cover middleware, mobile, database, container, AI, big data and many other fields, including Apache Dubbo, which has graduated from the Apache Foundation

First, Jia Yangqing led Ali Open Source, a blockbuster release developer cloud

Jia Yangqing use

Three isCloud primary:At present, there are many technologies designed on the cloud, and Ali Open Source will provide developers with a better and more accessible cloud native environment.

Operating system:Operating system is the base of the whole computer technology. Ali has the base of Flying Operating System. At the same time, he has applied many open source systems like Linux itself.Front end:Front-end is an important medium for interaction with users. To do a good job of front-end can quickly hand over technology to users. Ant Design of Ant Gold Clothes is one of the representative projects that have done very well.

So, what are we going to do for the open source community itself? Jia Yangqing said that we musttoolDo a good job, to provide you with more detailed, tested methodology, to help developers more easily promote their own open source projects. How to use open source projects and solutions to improve development efficiency, including test deployment tools, source code management tools, project communication platform, etc.

When it comes to developing caffe, at Berkley and participating in the core development of open source frameworks such as TensorFlow,pytorch, Jia Yangqing says all this is

The global open source community is booming, and in 2018, it grew faster than the previous years combined.

Open source is from 0 to 1, but one obstacle to open source projects is landing: how to actually find a landing scenario to support further development, so that open source projects can be implemented from 1 to 100. Ali provides a technical testing ground.

Take the in-depth learning framework as an example. It was first developed in the laboratories of Caffe, Torch and Theano. Later, with more and more industrial applications, more mature frameworks such as TensorFlow and Pytorch were developed. Today, open source software is more closely integrated with business.

Looking back, software like TensorFlow and Pytorch is a bit like assembly language in actual modeling. But what is the high-level language in the field of AI and big data? Should it stay in assembly language? Jia Yangqing believes that no, there are many frameworks for modeling, and the combination of large data frameworks is the way forward.

Jia Yangqing said that with the accelerated arrival of industry data intelligence, Aliyun's large-scale computing capabilities and rich practice scenarios will help reduce open source costs and promote the maturity and perfection of the project. Taking Apache Flink as an example, a large number of new functions have been accumulated in stream computing and batch processing through the field test of multiple core scenarios such as double 11 and search recommendation, and the performance has been improved rapidly. Today, Apache Flink has become the first choice for CCTV, Sina Weibo, Qianxuan location and other well-known companies to build stream processing platforms.

Today, Ali has also brought a blockbuster release:Developer cloud。 It will provide developers with zero threshold cloud experience, best hardware, cloud ecosystem, etc.

From embracing open source to contributing to open source, independent open source, and now enabling open source, open source has been upgraded to one of Ali's technology strategies, which will bring more tangible and far-reaching value to developers.

2. From Caffe, TensorFlow to PyTorch, Jia Yangqing's AI Architecture Goddess Road

When it comes to Jia Yangqing, we can't ignore Caffe and Tensor Flow.

Jia Yangqing established his AI status in Silicon Valley because of two world-class frameworks, Caffe and TensorFlow.

Caffe is the first general deep learning framework in history and one of the most widely used deep learning frameworks in the world.Microsoft, Yahoo, Nvidia, Adobe and other world-class companies are using it.

TensorFlow, had to mention the robot AlphaGo, which became famous for go (AlphaGo), TensorFlow is the powerful support behind AlphaGo, known asGoogle brain

Jia Yangqing is one of the authors of these two world-class frameworks.


Jia Yangqing LinkedIn Home Page

After graduating from Tsinghua Benshuo, Jia Yangqing received his doctorate in computer science from the University of California, Berkeley. Jia Yangqing developed an in-depth learning framework during his study in Berkeley.CaffeFull nameWith excellent structure, performance and code quality, it has become one of the most popular and successful open source in-depth learning frameworks so far.It has greatly promoted and influenced the development of the field. After graduating from Ph.D., Jia Yangqing joined Google, where he had been interning for two years before joining, and worked under Jeff Dean, the chief architect of Google, during which he participated in the development of TensorFlow platform. Jeff Dean is known as the leading figure in the whole of Google, Silicon Valley and computer science.

From 2014 to 2016, Jia Yangqing's academic outbreak occurred. According to Google Academic Statistics, Jia Yangqing's papers are quoted as high asThe 48560 timeOf whichGoing deeper with convolutionsHttps://www.cs.unc.edu/~wliu/papers/Google LeNet.pdf

The second highly cited paper is about Caffe:Caffe: Convolutional architecture for fast feature embeddingThe number of citations reached 11598. Paper address:Https://arxiv.org/pdf/1408.5093.pdf

The third highly cited paper is related to Tensorflow:Tensorflow: A system for large-scale machine learningThe number of citations reached 9266. Paper address:Http://download.tensorflow.org/paper/whitepaper2015.pdf


Jia Yangqing joined Facebook in February 2016 on Google Academic Page, and the news also sparked widespread discussion in the circle. See Jia Yangqing, Caffe's author of Xin Zhiyuan's report:Why did I leave Google and join Facebook?

Since then, Jia Yangqing has marched all the way along the road of AI architecture:

From the in-depth study of open source frameworks, to the current contest between TensorFlow and PyTorch, Jia Yangqing has also moved from Caffe author to Facebook AI Architecture Director.

He co-led the PyTorch 1.0 project, created the first prototype of open model format in 2017, and later released it with FB, Microsoft, Amazon and many hardware vendors.ONNX

In March this year,Jia Yangqing was revealed to have left Facebook and joined AlibabaHe is currently Vice President of Alibaba, General Manager of Alibaba Intelligent Computing Platform Business, and Head of Alibaba Open Source Technical Committee.

3. How did scientists grow up when they got rid of the dust and spoke faster than Rapper?

Inside Ali, people call Jia YangqingScientists Speaking faster than Rapper

Jia Yangqing was born in Shangyu, Shaoxing. His mother and father are both Chinese teachers in Shangyu Chunhui Middle School.

In the past few years, Shaoxing Daily reported on the growth experience of gifted teenagers in their hometown as follows:

When Jia Yangqing was over one year old, he liked to listen to stories. His mother told him stories with picture books. At the age of three or four, Jia Yangqing was already literate for two or three hundred years. He liked to read quietly with a book in his hand.

Once, a family of three went to visit Xinhua Bookstore,5 years old Jia Yangqing picked a text version of the "Andersen fairy tale ". The mother asked:

When I was in primary school, because of the inconvenience of transportation, the couple thought twice and decided to send their son to a village near the school. This village has double education, that is, the first and second grade children attend classes together.

Before going to elementary school, Jia Yangqing's mother earnestly told her son a few things by writing home letters:

In the middle school entrance examination, Jia Yangqing entered Chunhui Middle School with the third place in the middle school entrance examination in Shangyu District. Up to high school, my grades were absolute, and I got them.First Prize in National Physics and Chemistry Competition, Second Prize in Mathematics League and Second Prize in Comprehensive English AbilityWait.

As a learning hegemony, Jia Yangqing has his own learning routine. For example, for the questions assigned by the teacher, we will not do them once we know them, but we will buy a whole exercise until we are familiar with them.

In the college entrance examination, because of carelessness, Jia Yangqing made a mistake in a 27-point comprehensive examination. At one time, he was very frustrated, but at last, he took the comprehensive examination as an example.The score of 686 points in College Entrance Examination went smoothly into Automation Major of Automation Department of Tsinghua University

The mother always reminded his son of his success:

As the developer of Caffe, an open source software for image recognition, Jia Yangqing shared this achievement free of charge to the whole world, so his name is well known in the industry.

Finally, to understand Ali Kaiyuan, please see this picture with a lot of information.


A Picture to Understand Alibaba Open Source

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