From Microsoft Research Asia to Jinshan, and then retiring last week, Dr. Zhang Hongjiang can say that he has witnessed the development of China's corporate service market. Recently, he also came to the 36-year Wise2016 Enterprise Service Annual Meeting and won the Enterprise Service Lifetime Achievement Award on the spot. At the same time, Dr. Zhang Hongjiang also shared some of his views on the future of corporate services.
In terms of big data, Zhang Hongjiang takes AlphaGo as an example. Its training data far exceeds the data of the human master for a lifetime. Together with its computing power of more than 900 CPUs and 2,800 GPUs, its segment is very close to the human master. . "This is before the game, without considering other factors in it, its data is already very advanced." And in addition to Go, global data increased at a rate of 40%, Wal-Mart will produce 2.5PB every 4 hours. The amount of data, Twitter has 500 million posts updated every day.
Regarding whether AI will replace humans, Dr. Zhang Hongjiang also gave a similar view to Li Kaifu. The application of artificial intelligence is limited. It must have five conditions: big data, clear boundary, external feedback, top data scientists and computing resources.
Zhang Hongjiang believes that the application of AI+ big data will become the standard of enterprises. “In the past, it was analysis, now is prediction, and the future is cognition.†In terms of AI and data, it is inseparable from cloud computing, and it has become the mainstream technology of IT. Cloud has also become the fastest growing 2B industry.
What should a startup do after giving AI+ big data + cloud = the future of corporate services? Dr. Zhang Hongjiang also gave suggestions:
AI+ Big Data + Cloud will be with all new technologies, and the business opportunities will be fleeting. Entrepreneurs need a forward-looking vision to quickly accumulate our technical capabilities, engineering capabilities, and technical barriers that require craftsmanship. , can continuously improve our algorithm and analysis capabilities, and provide better services for enterprises. The following is a speech by Dr. Zhang Hongjiang:
The three stages of the enterprise service I am talking about today are closely related. It can be called 3.0+ or ​​Enterprise Service 4.0. If we recall what themes in 2016 are always running through the Internet, the entire enterprise service software market, and the entire IT market, there must be three themes: artificial intelligence, big data, and cloud computing.
Why do you say that?
Artificial intelligence is an area I have been doing for many years. Today, I talked about artificial intelligence. One of the big things in 2016 was AlphaGOvs Li Shishi in March. Many people don’t know what the word AI is, and I don’t know how much AI is. Great, but AlphaGo won the Korean Go Master, which is a milestone.
Almost 20 years ago, AI won the Russian chess master. In the past 20 years, artificial intelligence has gone a long way. After this incident, many people think that artificial intelligence will rule the world. I don't know if anyone here knows what artificial intelligence is all about before this game. Does anyone have a gamble with your colleague, whether it is a machine win or a person can win, I gambled and won the money. .
There is a very important reason for this. Inside, why do I think that AlphaGo can defeat the master of Go? It is very important that the deep support behind it. In fact, today, there is enough computing power. Today, there is enough data today. Artificial intelligence algorithm (machine learning algorithm).
I especially want to emphasize that in the data and algorithm, the human master can put thousands of real high-quality games in his life, but AlphaGo has been in the middle of its limited life and the human 6-9 players. 16,000 times, and 30 million machine matches, because it has no so-called emotional factors, so you can follow yourself, so these two data make its training data far exceed the data of the human master for a lifetime.
Coupled with its very strong computing resources, more than 900 CPUs, 2800 GPUs, and its segment is very close to the human master. This is before the master game, did not consider other factors, that is to say, its data volume is much ahead of the human master.
So the real power behind the AlphaGo is actually what we call big data and strong calculations plus new algorithms. This is the real reason behind the hottest thing that happened this year.
So when we talk about big data, we talked a lot in the middle of the past few years. Some people have asked whether the big data era is a question that has already passed. It is not, but it has just begun, and with the development of artificial intelligence, big data Applications will be more and more extensive.
In the past few years, with the development of mobile Internet, human data is actually exploding. According to IDC survey, the data created by human beings is about 4.4Z. By 2020, this number will increase by another 10 times. If you use A standard Apple Ipad with 128G of memory inside. If you deposit these 44 Zs, you can go back and forth 9 times from Earth to Moon.
For example, Wal-Mart's data generated every 4 hours in 2013 reached 2.5PB. Today, Twitter pushes 500 million words every day. Wechat shares 1 billion photos in a circle of friends. Today, China's stock photo is 30 billion. Zhang, but the data growth in the corporate world is also very fast.
I still use the mobile Internet example. Nowadays, the most common application on mobile phones is photos. After more and more photos, it is a problem to find management photos on your mobile phone. So most of the mobile phones began to appear called photos two years ago. Automatically sorted cloud services, including categorization based on faces. When you see these photos, you can only sort them by location and time. Today you can know how to classify your friends and family based on people's faces. . If there is a photo of Lei Jun on my mobile phone, I will click on his photo. You will find all the photos he has appeared to help you find out. The technology behind this is the artificial intelligence face recognition technology. This technology is today. It has appeared on all the first-line platforms.
This matter has been going on for decades in the history of AI. It is based on my patent. This is the patent I applied for at the HP Lab in 1997. The basic core of this patent is: I will make a face when the photo comes in. Detection, face feature extraction, and thus comparison, and I predicted that this system is difficult to implement on a system 20 years ago, but requires a distributed system, so it is a distributed database system, today It seems that this thing is a mobile phone.
This patent was approved in 2000. But this thing has been at least 20 years since I saw it. It was very difficult to do 20 years ago. Except that the algorithm is not as advanced as it is today, what is more important is that our data is not enough. So in my opinion for more than 20 years, the development of computer face recognition is a database and computing power expansion. In the early 1990s, the training data you used was hundreds of people and dozens of people. By the time of ten years ago, there were tens of thousands or hundreds of thousands of copies. The number of people we can use is also several thousand. It is really in the past. In the past three or five years, computers or artificial intelligence began to exceed the number of people. A typical situation is that Google and Facebook started using a large amount of data in early 2015, including 200 million photos, 8 million people, because there are Multi-data can support a very large network of deep learning, so that algorithms and face recognition systems with recognition accuracy exceeding 99.63% can be trained.
When we used 2.6 million training data, the recognition rate was about 76%. When we used 26,000, the recognition rate increased by 10%. You can see the data without careful tuning. The impact on recognition accuracy is also the same between the complexity and computational power of the algorithm you use, and can only be supported by increasing computational complexity and precision. My point is that data and computing power are the main drivers behind AI.
In the past two years, Chinese companies have sprung up in this area, leading the international market and leading the international academic field. These companies derived from Microsoft Research have achieved the best face recognition system in the world. The reason is that they are all with Microsoft Research. Perhaps more importantly, they all have a lot of data today. China has the largest population, China has the largest number of cameras and the highest density of cameras. This series of data allows us to make the best recognition system.
In addition to the human face, we know that as long as you have done a bad thing, as long as you take a picture of you in any corner of the road, you basically can't escape. In addition to face recognition, AI will exceed in any field. Human beings, this is the core content we are talking about today. AI has replaced manpower in a series of fields, besides face recognition, chess, translation, journalist editing, especially the editing of financial journalists, in administrative assistants. Police, taxis, stock traders, accountants, including nannies will be replaced by AI to a large extent in the next 5 to 10 years.
why? You ask yourself a few questions. Can you count millions of flags every day like AlphaGo? Can humans collect data from more than 200,000 cars on the road every day like Tesla, can you like all of them? There are so many people in the airport and the cameras in the train stations. It is precisely because of these endless data that AI has more and more powerful capabilities, so that they are not only able to recognize people, but also understand more languages ​​than humans.
So we also believe that all of this will change the shape of our future corporate services. You will ask a question, is AI able to replace all industries, all types of work, all tasks, it is very important that you must be able to do a good AI system with big data, there must be a dial A good top scientist can make this data a good system, so that there are good algorithms and a lot of computer resources.
In fact, we also know here that whether you are a computer resource or a scientist, you can find it. For startups, big data is a challenge for startups, and you can get so much data.
Everyone will ask a question, how much data is enough? In the analysis of big data, we never have too much data, the data is never enough, because it is very important. In the traditional AI way, the performance is also increasing with the increase of the data. The number added by the latest algorithm will be faster, or the saturation speed will be slower. This is very important. When you don't have data, you can use the latest algorithm. Your performance will only be here. It won't be here, especially the more and more complicated scenes we encounter today. What you need The data will be more and more. The important thing is that when your coverage and data accuracy reach a certain level, the dependence on the traditional model will be reduced, and a model will be made when the data is lacking.
Today, we look at the impact of big data on enterprises. I want to use the data of a survey in the United States. The application of big data has been very extensive. In 60% of IT companies, big data has become the daily middle of use, in business and professional. Nearly half of the companies in the service company are in use, 47% of the financial companies, and the most traditional manufacturing and retail industries are gradually rising, so we know that in our corporate service industry today, the data has become The next new religion.
In the past, traditional reliance on big data used it to do statistics. We know what happened in the past. We use it for analysis very quickly. We want to know why this happened. Today, more people are using forecasts, what will happen in the future. In the future, I think we will use it. The biggest one is cognition. Our feedback on any business decision, from simple knowledge to intelligent evolution, is one of the big data that today plays a role in enterprise services. The core reason.
When you have big data, advanced algorithms, like face-to-face demonstrations, face recognition on your phone, you might think that the phone itself has such powerful features, but I want to tell you that all the calculations are done in the cloud. After your photo is taken, it is sent to the cloud, the face is detected and recognized in the cloud, the contact's database is searched in the cloud, and the information of the face of the photo is returned, that is, if There is no back-end, no large-scale storage in the cloud, large-scale computing, large-scale data analysis, a series of computing functions, today can not do what we say.
Cloud computing has become a mainstream technology. Today, unlike three years ago, when you talk about cloud computing, others will think that you are talking about it in the fog. Today cloud computing has become the real mainstream of IT. To start a company today, you need When you have a lot of computing resources, what you think of is not to buy a data center yourself, but to buy a lot of servers, but to take advantage of the cloud computing providers that a large number of cloud services providers already have.
China's cloud computing capacity is very broad and the potential is unlimited. Because I share another data, IDC made a report on China's public cloud market at the end of last year. China's public cloud market, including SaaS, IaaS, all public ownership. The total market revenue of the cloud accounts for only 3% of the total income of the US. But in which field of China's Internet market, we are only more than 3% of the US? Either the same grade as the United States, or bigger than the United States. Even if it is smaller, we are at least half of the United States. When you see a rapidly growing market, when the entire scale of the market is only 3% of the US, you know how big this market is, so China's cloud market has unlimited potential.
Finally, I would like to summarize that when we talk about enterprise services, especially when it comes to enterprise services in the new era, there are three points we must fully consider:
First of all, artificial intelligence, big data and cloud computing capabilities will be like any previous technology. Any new technology will appear, and its opportunities will be fleeting. For the phenomenal technology explosions such as AI, big data and cloud, What is needed is a very forward-looking vision, the ability to follow the trend; we need to quickly accumulate our technical and engineering capabilities; we also need to have the spirit of craftsmanship, able to continuously improve our algorithms and analytical capabilities, so that we can Really provide enterprises with the best new era of corporate services. thank you all!
Zhang Hongjiang: AI+ Big Data + Cloud = The Future of Enterprise Services
The first is artificial intelligence. Dr. Zhang Hongjiang shared his interesting story of winning money during AlphaGoVS Li Shishi. This world-famous competition also gave the public a preliminary understanding of the capabilities of AI. "Almost 20 years ago, artificial intelligence The machine won Russia's chess, but in the past 20 years, artificial intelligence has gone a long way." Dr. Zhang Hongjiang said that behind this is actually the support of big data, new algorithms and strong calculations.In terms of big data, Zhang Hongjiang takes AlphaGo as an example. Its training data far exceeds the data of the human master for a lifetime. Together with its computing power of more than 900 CPUs and 2,800 GPUs, its segment is very close to the human master. . "This is before the game, without considering other factors in it, its data is already very advanced." And in addition to Go, global data increased at a rate of 40%, Wal-Mart will produce 2.5PB every 4 hours. The amount of data, Twitter has 500 million posts updated every day.
Regarding whether AI will replace humans, Dr. Zhang Hongjiang also gave a similar view to Li Kaifu. The application of artificial intelligence is limited. It must have five conditions: big data, clear boundary, external feedback, top data scientists and computing resources.
Zhang Hongjiang believes that the application of AI+ big data will become the standard of enterprises. “In the past, it was analysis, now is prediction, and the future is cognition.†In terms of AI and data, it is inseparable from cloud computing, and it has become the mainstream technology of IT. Cloud has also become the fastest growing 2B industry.
What should a startup do after giving AI+ big data + cloud = the future of corporate services? Dr. Zhang Hongjiang also gave suggestions:
AI+ Big Data + Cloud will be with all new technologies, and the business opportunities will be fleeting. Entrepreneurs need a forward-looking vision to quickly accumulate our technical capabilities, engineering capabilities, and technical barriers that require craftsmanship. , can continuously improve our algorithm and analysis capabilities, and provide better services for enterprises. The following is a speech by Dr. Zhang Hongjiang:
The three stages of the enterprise service I am talking about today are closely related. It can be called 3.0+ or ​​Enterprise Service 4.0. If we recall what themes in 2016 are always running through the Internet, the entire enterprise service software market, and the entire IT market, there must be three themes: artificial intelligence, big data, and cloud computing.
Why do you say that?
Artificial intelligence is an area I have been doing for many years. Today, I talked about artificial intelligence. One of the big things in 2016 was AlphaGOvs Li Shishi in March. Many people don’t know what the word AI is, and I don’t know how much AI is. Great, but AlphaGo won the Korean Go Master, which is a milestone.
Almost 20 years ago, AI won the Russian chess master. In the past 20 years, artificial intelligence has gone a long way. After this incident, many people think that artificial intelligence will rule the world. I don't know if anyone here knows what artificial intelligence is all about before this game. Does anyone have a gamble with your colleague, whether it is a machine win or a person can win, I gambled and won the money. .
There is a very important reason for this. Inside, why do I think that AlphaGo can defeat the master of Go? It is very important that the deep support behind it. In fact, today, there is enough computing power. Today, there is enough data today. Artificial intelligence algorithm (machine learning algorithm).
I especially want to emphasize that in the data and algorithm, the human master can put thousands of real high-quality games in his life, but AlphaGo has been in the middle of its limited life and the human 6-9 players. 16,000 times, and 30 million machine matches, because it has no so-called emotional factors, so you can follow yourself, so these two data make its training data far exceed the data of the human master for a lifetime.
Coupled with its very strong computing resources, more than 900 CPUs, 2800 GPUs, and its segment is very close to the human master. This is before the master game, did not consider other factors, that is to say, its data volume is much ahead of the human master.
So the real power behind the AlphaGo is actually what we call big data and strong calculations plus new algorithms. This is the real reason behind the hottest thing that happened this year.
So when we talk about big data, we talked a lot in the middle of the past few years. Some people have asked whether the big data era is a question that has already passed. It is not, but it has just begun, and with the development of artificial intelligence, big data Applications will be more and more extensive.
In the past few years, with the development of mobile Internet, human data is actually exploding. According to IDC survey, the data created by human beings is about 4.4Z. By 2020, this number will increase by another 10 times. If you use A standard Apple Ipad with 128G of memory inside. If you deposit these 44 Zs, you can go back and forth 9 times from Earth to Moon.
For example, Wal-Mart's data generated every 4 hours in 2013 reached 2.5PB. Today, Twitter pushes 500 million words every day. Wechat shares 1 billion photos in a circle of friends. Today, China's stock photo is 30 billion. Zhang, but the data growth in the corporate world is also very fast.
I still use the mobile Internet example. Nowadays, the most common application on mobile phones is photos. After more and more photos, it is a problem to find management photos on your mobile phone. So most of the mobile phones began to appear called photos two years ago. Automatically sorted cloud services, including categorization based on faces. When you see these photos, you can only sort them by location and time. Today you can know how to classify your friends and family based on people's faces. . If there is a photo of Lei Jun on my mobile phone, I will click on his photo. You will find all the photos he has appeared to help you find out. The technology behind this is the artificial intelligence face recognition technology. This technology is today. It has appeared on all the first-line platforms.
This matter has been going on for decades in the history of AI. It is based on my patent. This is the patent I applied for at the HP Lab in 1997. The basic core of this patent is: I will make a face when the photo comes in. Detection, face feature extraction, and thus comparison, and I predicted that this system is difficult to implement on a system 20 years ago, but requires a distributed system, so it is a distributed database system, today It seems that this thing is a mobile phone.
This patent was approved in 2000. But this thing has been at least 20 years since I saw it. It was very difficult to do 20 years ago. Except that the algorithm is not as advanced as it is today, what is more important is that our data is not enough. So in my opinion for more than 20 years, the development of computer face recognition is a database and computing power expansion. In the early 1990s, the training data you used was hundreds of people and dozens of people. By the time of ten years ago, there were tens of thousands or hundreds of thousands of copies. The number of people we can use is also several thousand. It is really in the past. In the past three or five years, computers or artificial intelligence began to exceed the number of people. A typical situation is that Google and Facebook started using a large amount of data in early 2015, including 200 million photos, 8 million people, because there are Multi-data can support a very large network of deep learning, so that algorithms and face recognition systems with recognition accuracy exceeding 99.63% can be trained.
When we used 2.6 million training data, the recognition rate was about 76%. When we used 26,000, the recognition rate increased by 10%. You can see the data without careful tuning. The impact on recognition accuracy is also the same between the complexity and computational power of the algorithm you use, and can only be supported by increasing computational complexity and precision. My point is that data and computing power are the main drivers behind AI.
In the past two years, Chinese companies have sprung up in this area, leading the international market and leading the international academic field. These companies derived from Microsoft Research have achieved the best face recognition system in the world. The reason is that they are all with Microsoft Research. Perhaps more importantly, they all have a lot of data today. China has the largest population, China has the largest number of cameras and the highest density of cameras. This series of data allows us to make the best recognition system.
In addition to the human face, we know that as long as you have done a bad thing, as long as you take a picture of you in any corner of the road, you basically can't escape. In addition to face recognition, AI will exceed in any field. Human beings, this is the core content we are talking about today. AI has replaced manpower in a series of fields, besides face recognition, chess, translation, journalist editing, especially the editing of financial journalists, in administrative assistants. Police, taxis, stock traders, accountants, including nannies will be replaced by AI to a large extent in the next 5 to 10 years.
why? You ask yourself a few questions. Can you count millions of flags every day like AlphaGo? Can humans collect data from more than 200,000 cars on the road every day like Tesla, can you like all of them? There are so many people in the airport and the cameras in the train stations. It is precisely because of these endless data that AI has more and more powerful capabilities, so that they are not only able to recognize people, but also understand more languages ​​than humans.
So we also believe that all of this will change the shape of our future corporate services. You will ask a question, is AI able to replace all industries, all types of work, all tasks, it is very important that you must be able to do a good AI system with big data, there must be a dial A good top scientist can make this data a good system, so that there are good algorithms and a lot of computer resources.
In fact, we also know here that whether you are a computer resource or a scientist, you can find it. For startups, big data is a challenge for startups, and you can get so much data.
Everyone will ask a question, how much data is enough? In the analysis of big data, we never have too much data, the data is never enough, because it is very important. In the traditional AI way, the performance is also increasing with the increase of the data. The number added by the latest algorithm will be faster, or the saturation speed will be slower. This is very important. When you don't have data, you can use the latest algorithm. Your performance will only be here. It won't be here, especially the more and more complicated scenes we encounter today. What you need The data will be more and more. The important thing is that when your coverage and data accuracy reach a certain level, the dependence on the traditional model will be reduced, and a model will be made when the data is lacking.
Today, we look at the impact of big data on enterprises. I want to use the data of a survey in the United States. The application of big data has been very extensive. In 60% of IT companies, big data has become the daily middle of use, in business and professional. Nearly half of the companies in the service company are in use, 47% of the financial companies, and the most traditional manufacturing and retail industries are gradually rising, so we know that in our corporate service industry today, the data has become The next new religion.
In the past, traditional reliance on big data used it to do statistics. We know what happened in the past. We use it for analysis very quickly. We want to know why this happened. Today, more people are using forecasts, what will happen in the future. In the future, I think we will use it. The biggest one is cognition. Our feedback on any business decision, from simple knowledge to intelligent evolution, is one of the big data that today plays a role in enterprise services. The core reason.
When you have big data, advanced algorithms, like face-to-face demonstrations, face recognition on your phone, you might think that the phone itself has such powerful features, but I want to tell you that all the calculations are done in the cloud. After your photo is taken, it is sent to the cloud, the face is detected and recognized in the cloud, the contact's database is searched in the cloud, and the information of the face of the photo is returned, that is, if There is no back-end, no large-scale storage in the cloud, large-scale computing, large-scale data analysis, a series of computing functions, today can not do what we say.
Cloud computing has become a mainstream technology. Today, unlike three years ago, when you talk about cloud computing, others will think that you are talking about it in the fog. Today cloud computing has become the real mainstream of IT. To start a company today, you need When you have a lot of computing resources, what you think of is not to buy a data center yourself, but to buy a lot of servers, but to take advantage of the cloud computing providers that a large number of cloud services providers already have.
China's cloud computing capacity is very broad and the potential is unlimited. Because I share another data, IDC made a report on China's public cloud market at the end of last year. China's public cloud market, including SaaS, IaaS, all public ownership. The total market revenue of the cloud accounts for only 3% of the total income of the US. But in which field of China's Internet market, we are only more than 3% of the US? Either the same grade as the United States, or bigger than the United States. Even if it is smaller, we are at least half of the United States. When you see a rapidly growing market, when the entire scale of the market is only 3% of the US, you know how big this market is, so China's cloud market has unlimited potential.
Finally, I would like to summarize that when we talk about enterprise services, especially when it comes to enterprise services in the new era, there are three points we must fully consider:
First of all, artificial intelligence, big data and cloud computing capabilities will be like any previous technology. Any new technology will appear, and its opportunities will be fleeting. For the phenomenal technology explosions such as AI, big data and cloud, What is needed is a very forward-looking vision, the ability to follow the trend; we need to quickly accumulate our technical and engineering capabilities; we also need to have the spirit of craftsmanship, able to continuously improve our algorithms and analytical capabilities, so that we can Really provide enterprises with the best new era of corporate services. thank you all!
Our factory can produce the galvanized steel poles with various heights.
Steel Lighting Pole,Steel Lamp Post,Metal Lamp Post,Metal Pole For Outdoor Lights
Jiangsu Baojuhe Science and Technology Co.,Ltd. , https://www.galvanizedsteelpole.com