Text: Zhang Ying, Ye Kunkun, Deng Cong, Liu Tong
Source: People's Post and Telegraph
Editor's Note : On December 22, the "Wind Direction Series of Events and He Baohong Thought Enjoyment", co-sponsored by the People's Post and Telegraph Press and People's Posts and Telecommunications Press, was held in Beijing. At the meeting, He Baohong, director of the Institute of Cloud Computing and Big Data of the China Academy of Information and Communications Technology, used his unique perspective to look beyond technology and use humorous, humorous, and vivid language to tell the deepest laws behind technological development. Blockchain, artificial intelligence, big data, cloud computing, open source … Let's count the technical outlets that must not be missed in 2020.
- Observations | Blockchain and Artificial Intelligence: A catalyst for the “fourth industrial revolution”?
- AI+ blockchain, Jarvis+ allows each community to have a personal assistant
- Blockchain and the future of the intelligent revolution
- 44 "blockchain + AI" application scenario analysis, do you think it can be more complete?
- AI+Internet of Things + Blockchain: These 10 companies need to be medical network security guardians
- It is not enough to have a blockchain, and there is a way out in combination with AIOT.
Facing the upcoming 2019, several major industries have experienced different ups and downs. In 2019, the Internet industry experienced a bit of coolness. Virtual reality has made a more rational choice. It seems that the martyrs are more than the pioneers in retreating to the entertainment and game markets. Facing the two mountains that are difficult to climb, the left hand is to protect privacy, and the right hand is asset flow; artificial intelligence faces seven years of itching; the blockchain has experienced a 10-month trough and a two-month climax. At the same time, the Internet industry has sent warmth, and the industrial Internet is booming; 5G is the most dazzling technology; "Video +" makes everything video; "Cloud +", the cloud was looking for you in the past, and now it is You look for the cloud; in addition, edge computing, data, intelligence, servers, data centers, etc., have all experienced more or less changes and developments in 2019. For more than a decade, every few days, a technology has been announced that will disrupt the world. Behind change is eternity. Only by exploring eternity can we grasp change.
What is changing? For decades, the Internet has been undergoing transformation and development. It is shifting from the consumer Internet to the industrial Internet. The consumer Internet dividend has been basically "eaten up", but after five years of hard work, the industrial Internet dividend has not yet arrived. The dividends of the mobile Internet have basically been "eaten up" and are developing towards the Internet of Things and wearable devices. Similarly, the dividends of IoT and wearable devices have not yet arrived. From computing products to cloud computing, "vision" has shifted from human vision to machine vision, and is mainly used in areas such as unmanned driving and video surveillance. Data is increasingly becoming an asset from information flow.
After more than ten years of transformation and development of the Internet, this transformation and development has some different "flavors"-two "clouds" float in the distant sky. First, Moore's Law provides a steady stream of free or cheap computing resources for the information and communication industry, but when the resources on which the entire industry relies no longer make it abundant and cheap, it will face new challenges. Second, after years of development in the industry, the pace of technological innovation is too fast, and it is unsustainable to face the status quo that the pace of technological application is far behind the pace of technological innovation.
No matter where we go, human beings are moving from industrial civilization to digital civilization. As a new technological species, the Internet's inversion of the external living environment will certainly change some effective methods and some unbreakable "truths" for half a century.
The internet of value cannot replace the internet of information
With the deep integration of the Internet and the real economy, data increasingly has asset attributes. At the same time, data is information, which must be disseminated, open and shared. When data represents assets, the same network system is needed to transfer it. The value Internet cannot replace the information Internet, because when data is used to represent assets, it is still information. Therefore, the future value Internet will be built on the information Internet, and it is impossible to replace the information Internet. How to understand blockchain from the perspective of data assets? The last owner of the data is the owner of the corresponding asset. The previous owners of the data are only the owners of the corresponding information. The value of the owner on the blockchain is only to prove who is the last owner of the data.
What's new in blockchain? It requires multiple parties to maintain, traditional databases need only one party to maintain. Traditional databases belong to a certain enterprise, individual or organization, and need a certain entity to maintain the data of the database. There are multiple parties involved in the blockchain, not just a certain entity or individual, which requires the parties to jointly maintain. Blockchain technology involves cryptography, distributed consensus, smart contracts, etc., and is not easy to tamper with.
Decentralization is a pseudo-proposition, shouting decentralization, mostly because you want to become a new center. The consequence of decentralization is that the blockchain will become a new center.
The current state of blockchain technology can be summarized in six words: it is available and not easy to use. There is no problem with the technical direction, but there are some areas that need improvement. For example, change from lone chain to cross-chain, combine from on-chain to on-chain, off-chain, cut the whole chain into small blocks, and so on. In addition, the topological structure is being optimized, and trees, bridges, graphics, and dots have appeared. Although the name is still called blockchain, it is no longer a chain structure.
Blockchain is not a panacea and will not subvert the world. It is a new type of database. When the database is used, the blockchain is considered. It requires multiple agents. Everyone must participate together to solve the problem of mutual distrust in the absence of a third party.
When to use a public chain and when to use an alliance chain? In many cases, multiple parties trade through the market. In order to protect business secrets, as long as the participants know each other, the issue of confidentiality must be considered at this time, and the alliance chain or private chain must be used.
The blockchain is known as a "trusted machine", but it has become a scam tool many times now. Why? The reason is simple, trust is lacking and supervision is needed. When can I trust the data of the blockchain? When the blockchain data is native, you can use it with confidence; on the other hand, if the data is ever exposed, you need to be very careful. A lot of data has been tampered with before you see it. In addition to the data, we also need to look at the underlying system to prove that the set of machines that create trust is itself trustworthy. This is the reason and value of the trusted blockchain.
New Infrastructure for Digital Society
The data center is the infrastructure of the entire digital industry infrastructure. Regardless of whether it is global or China, the growth of data centers as infrastructure infrastructure is very fierce. Data centers, like computers, are at two extremes. On the one hand, data centers are becoming smaller and smaller on the consumer side, from PCs to handheld devices, to wearable devices to the Internet of Things; on the other hand, toward high-performance computing, data centers are becoming larger and larger. Global data grows by about 50% each year, while data centers grow by 20% to 30%.
One of the trends in data center development is productization. In the past, the construction of a data center was more like a construction site, and the site construction was far greater than the productization. The core of the Scorpio plan launched in 2011 is to pre-engineer, modularize, and prefabricate IT equipment, push it to the factory, and make construction on the site less and less. The Panama project launched this year has pushed products from IT to non-IT, and to power supply and distribution. Obviously, this road will go further and further. In the past few years, the server market has undergone tremendous changes. In the past, the computer room environment was often modified to adapt to standard servers. Telecom operators put forward new ideas to transform servers, and OTII was born. OTII is now a custom server to accommodate traditional communications bureaus.
Liquid cooling technology is becoming more and more mature, but the supporting external environment still needs us to maintain together. The increase of the calculation density of liquid cooling monomers by more than 10 times can reduce the area of IT equipment by more than 75%. HPC and GPU have already been applied in large scales, which can meet the needs of PUE construction in first-tier cities.
White box switches are rapidly rising; in the world of cabinets, light advances and copper retreats …
Network innovation is another main battlefield. Today, the focus of network innovation is mainly in two areas, one near the user side, one near the service side, and 5G near the user side, and data centers near the service side. It can be said that the data center is the commanding height of today's network innovation.
All in all, the data center is the new infrastructure of the digital society. The data center is the infrastructure of infrastructure such as the Industrial Internet, artificial intelligence, and the Internet of Things.
Almost impossible to hide
Open source is the foundation of social collaboration. As software becomes more expensive and riskier, cooperation within an enterprise alone is no longer sufficient. Global supply chains and global production chains are needed to jointly produce software, maintain software, and deliver software. Open source is also needed for social production. After 20 years of development, open source is not what it used to be. Twenty years ago, open source was a software development and delivery method; in 2020, open source had evolved into a model of ecological competition. The open source production and delivery model has matured. In 2019, 70% of the Fortune 50 companies contributed open source code to GitHub (code hosting website). Open source has formed two modes: foundation-led mode (Linux mode) and enterprise-led mode (Android mode). The number of code farmers has skyrocketed because code farmers are representative of advanced productivity in a digital society.
All in all, open source is almost inescapable, with 99% of organizations already using open source software in their IT systems. In 2019, China's code contribution increased by 48%. In China, the core of open source was the localization of international projects. Now, the focus of domestic open source is the localization of local projects. It is important to prioritize risk management to govern open source.
Big data leads us to believe that data is the oil of the 21st century. One of the inflection points of big data is the change from the "large" amount of data that was previously focused on to how to make data "run faster" today. The second inflection point of big data is to shift from focusing on data technology to now treating technology as an asset.
The two big mountains of big data-the left hand is how data is capitalized, and the right hand is how data is protected. The implementation of the GDPR for one year has triggered a global privacy protection boom. Regardless of whether it is the EU or China, relevant regulations for data privacy protection have been intensively issued. But the other direction is how to promote the controlled flow of data. Protecting privacy depends not on policies but also on technology. There are three schools of data protection policies so far: one is to desensitize the data and let the data "go out less"; the other is to secure multi-party calculations; the third is the federal calculation to let the data run less and the algorithm run more.
"Any new technology will create new problems while solving old problems." He Baohong said. The traditional data governance framework has failed. The fundamental reason is that data was information in the past, and today data is assets. It is obviously not applicable to move a set of information framework and methods to asset governance, so new data governance needs to be created. Framework, new data governance model. Data has changed from a cost center to a value center.
Data openness is not a matter of zeros and ones. There is a long gray area between the opening and closing of data, which needs to be solved through openness. Data opening requires data licenses and open data management methods. When it comes to personal privacy and user rights, the algorithm should be open. The open problem, neutral problem, and interpretability of the algorithm must be solved.
All in all, data is sometimes private, sometimes information, and sometimes assets, so an asset management approach is needed. "I believe that the technology about data management should become more and more fragmented in the future. This society needs more updated technologies to deal with different data."
The intelligence that computers can't yet achieve
After several years of development in the field of artificial intelligence, computer vision applications have thrived. Google could soon develop algorithms that modify natural language conversations and even predict jokes. Things that ordinary people can figure out in just one second will probably be automated by AI in the near future.
In 2019, AI has been itchy for seven years. Before 2015, AI was talent, technology, and capital; from 2016 to 2018, it was data and scenarios; in 2019, it began to focus on the issues of artificial intelligence ethics, morality, interpretability and regulation.
What kind of artificial intelligence does the industry need? Efficient, model training time needs to meet real-time business response; economical, to reduce computing costs; ubiquitous, to make AI computing everywhere; intelligent, automated semi-automatic data labeling; stable, to meet the needs of industrial AI systems
With the rise of video surveillance and driverless, artificial intelligence has developed to a new stage. Advances in artificial intelligence in recent years, such as face recognition and identity verification, have produced huge changes, but deep counterfeiting technology and deep counterfeiting must also be addressed.
He Baohong believes that there are pitfalls in the interpretability of AI. The first trap is that human intelligence is not interpretable; the second trap, 87% of people believe that their IQ is higher than the average; the third trap is unexplainable if it is unintelligible.
In short, all interpretable intelligence is computing, and artificial intelligence is intelligence that computers cannot yet implement.
Technology is spiraling
Industry needs to catch up with technological innovation
He Baohong believes that innovation is too fast and applications cannot keep up, and it will inevitably depart from the pace of industry innovation. Technology has always been circling and spiraling. This is because a technology has been moving in one direction for a long time, and its weakness in the opposite direction has been exposed. Just like cloud computing has been going for a long time, there must be edge computing. The Internet allows information and networks to be distributed, but SDN allows the network to be centralized, cloud computing to centralize computing, and big data to centralize data. For another example, the advantage of IT technology in the past was editable, but the advantage of blockchain now is not editable. Therefore, technology is spiraling. Today, three directions are of particular interest. The first is software. Everyone needs to work on software, especially parallel software. The second is the algorithm. Many of the things that everyone did before are adding. Almost all technical optimizations are basically how to complicate the technology. Today, we must find a way to do subtraction. The third is the chip, which used to be a general-purpose chip. In recent years, it is a special-purpose chip, especially a chip in a specific scenario, such as a special chip for edge computing and artificial intelligence. Obviously, there will be some big specific problems.
If 100 years ago, IT technology could not keep up with the development of the industry and society as a whole, today's problem is reversed. The whole society cannot keep up with the pace of IT technology innovation. This is definitely not sustainable. Now there are two opportunities. First, if the application can't keep up with the pace of innovation, then do the industry and application, the industrial Internet, and the combination of scenarios. The second opportunity is the construction of soft power. When new technologies are applied to various industries, economic, educational, institutional, and ethical impacts will be found. This requires optimization of economics, pedagogy, management systems, ethics, etc. Change applications, change society.