Netease Technology News December 8 news, 2018 NetEase economist annual meeting artificial intelligence forum, the International Artificial Intelligence Association Council Chairman, International Artificial Intelligence Association executive Yang Qiang gave a speech.
Yang Qiang said that he experienced two hot artificial intelligence. When the direction of the doctor was first confirmed, everyone urged him to choose artificial intelligence. Later, when he experienced artificial intelligence in winter, he also persisted. After that, the emergence of deep learning caused more and more exciting things in the artificial intelligence field.
How can artificial intelligence come to the ground and how does it have a huge impact? Yang Qiang believes that artificial intelligence is an empowering engine. It does not itself come with a commercial scenario. Therefore, when considering the landing of artificial intelligence, it must be equipped with a good business scenario. In Yang Qiang's opinion, it is good. The commercial scenario is that it has very clear boundary conditions, strong data characteristics, and there are large, high-frequency external feedbacks in this data. In addition, we must have strong computing resources behind support, but also have very advanced algorithms. Adding these together we will see a closed loop.
Yang Qiang believes that the closed loop of AI should have several characteristics. First of all, this closed loop should not involve anyone, but it must be automated. Second, it is this closed loop that generates such data, such feedback at a very high frequency, and finally such an iteration makes the function that produces performance must happen quickly, but cannot say that in a few years time, it must be several times. A period of weeks, or even days, sounds quickly. Artificial intelligence must go through a certain process. For example, we first need to define what is the artificial intelligence problem that we are facing. What is also particularly important is how good we can get after investing in artificial intelligence. We must analyze labor from an economic point of view. Smart landing, conduct a comparison. At the same time, we must have a short time, quickly find the data, select and train the model, and finally we can test the model.
Which artificial intelligence will be hot in the future? Yang Qiang believes that the first is Explainable AI's interpretable AI. Past deep learning is in the form of a black box. Explaining becomes important. For example, medical, financial, and government decisions require transparent interpretable AI. . The second is that AI for all lowers the threshold of artificial intelligence. In addition, there are migration studies, data that do not need to be built, and the acceleration of machine learning, which eventually makes AI a service.
Yang Qiang focused on his achievements in the transfer of learning. Migration learning, migration of existing models to new areas. Can solve small problems with small data to ensure personal privacy does not flow out. At present, Yang Qiang has designed fields such as machine reading, machine public opinion analysis, cross-domain multimedia knowledge transfer, and personalized dialogue systems.
Finally, Yang Qiang hopes that everyone can work hard to lower the technical threshold from the industrial perspective and realize the slogan of the fourth paradigm: AI for Everyone.
The following is the text of Yang Jiang's speech, slightly edited.
Yang Qiang: Hello, everyone. It is a great honor to meet you as the first speaker. Just like the host just said, artificial intelligence is now a very hot term, but how can artificial intelligence land on millions of households? This is a topic of special concern for each of us now.
First of all, let's take a look at the course of artificial intelligence. People like me at this age are older than everyone in the audience. I was fortunate to have experienced two artificial intelligence fluctuations. I remember the first time I contacted artificial intelligence in 1985. It was when I decided to go for the direction of a doctor. At that time, I was around. The alumni of these alumni all told me in unison to persuade me to read artificial intelligence because it was very hot. In the late 1990s, I experienced a cold winter. This was also the people we left behind. Then there was an exciting event that took place, for example, Deep blue defeated humans, such as the emergence of deep learning, so that everyone is familiar with AlphaGo, AlphaZero and so on. Let's take a look at these glorious events and we may be more concerned with how artificial intelligence really lands and can play its part in our lives and work. We also endeavor to analyze how artificial intelligence can be grounded, what conditions and what prerequisites make artificial intelligence really have a huge impact. One direction we look at is that artificial intelligence is an empowering engine. It does not itself come with a commercial scenario, but we must provide it with a good commercial scenario. What is a good commercial scenario? That is, it has very clear boundary conditions, has strong data characteristics, has big data, and there is a lot of high-frequency external feedback in this data. In addition, we must have strong computing resources behind support, but also have very advanced algorithms. Adding these together, we will see a closed loop. First of all, this closed loop is best not to have anyone involved, that is, automation. Second, it is this closed loop that generates such data, such feedback at a very high frequency, and finally such an iteration makes the function that produces performance must happen quickly, but cannot say that in a few years time, it must be several times. A period of weeks, or even days, sounds quickly. Artificial intelligence must go through a certain process. For example, we first need to define what is the artificial intelligence problem that we are facing. What is also particularly important is how good we can get after investing in artificial intelligence. We must analyze labor from an economic point of view. Smart landing, conduct a comparison. At the same time, we must have a short time, quickly find the data, select and train the model, and finally we can test the model. So such a process is essential for every artificial intelligence application.
Well, with this kind of thinking, we can also take a look and see. We now have such a result. What will be the hot spots in artificial intelligence in the future? If you have been to these more advanced academic conferences, you will find that there is a new term called "Explainable AI," which is now very popular. Why do you have such words? Because there may have been some explosions in the past, such as deep learning is in the form of a black box, we do not know how it works, so if you can explain the artificial intelligence engine is now becoming a very urgent task, such as in the key Landing scenes, decision-making, medical care, education, and the government’s economic decision-making all require such a “white box,†which is transparent and allows people to explain the reasons, causes, and consequences behind. There is also the ability of artificial intelligence to not lower the threshold, as the fourth paradigm is trying to do this kind of prophetic platform, so that ordinary people can also use artificial intelligence products to build applications on it. There is also unstructured data. This data is often in the form of natural language. It may appear in other signal forms. Because there is a large amount of artificial needs to clean up data, its progress is relative to other artificial intelligence. The field is relatively slow, but its application is more extensive. There is also how to make the artificial intelligence training process more agile, that is, how Speedup machine learning can flow. Finally, artificial intelligence can provide services to everyone. Just like we open tap water, we can get this service automatically. This makes the social division of labor clearer, so that people who do not know AI can get the benefits of AI. Can do this, that is, AI can not serve all people, we must study this technology, there is a technology that I and my students have been studying is called migration learning, migration learning that is how can It is possible to migrate an already reliable model and experience to a similar field, so that a good model can be obtained in a new field without spending so much resources. Why is migration learning useful? The first is that it can deal with small data, that is to say in a similar field, we can rely on small data in this field and big data in the previous field to create a new model in a new scenario of small data. The second is that it can solve many privacy issues. Suppose we want to migrate data to a personal terminal. If the terminal has the ability to adapt a generic model to a personalized model, it can quickly build a very reliable personalized model on the phone. There are news recommendations and the like, which can be used to this point. In the process of doing so, we can ensure that personal privacy does not flow out. This is very important in enterprise services. If a company establishes a model and serves another enterprise, the other enterprise is unnecessary and does not need to transfer the data to the previous enterprise. It can carry out model migration locally. This can make AI as a service available.
The work done in our lab, for example by letting the machine understand people's knowledge, allows the machine to conduct public opinion analysis, and after a model of public opinion analysis in one area can be migrated to related fields and be able to do cross-domain multimedia at the same time. For example, text-to-image knowledge transfer. There is also a dialogue system. If we have a general-purpose dialogue system, we can migrate to a personalized mobile phone for everyone. We can conduct dialogues, can recommend, and serve individuals.
Summary: The slogan of the fourth paradigm is the same as AI For Everyone. How can this be achieved? To really get this we need not only to learn, in the industry, to work hard to reduce the AI ​​threshold. If we can really do this, it is like a new big data and model-driven economic evolution. This is me. To say, thank you all.
Tin Zinc Antimony Copper Alloy Wire
It was made by Tin,Zinc, Antimony and Copper,Have a good metal characteristics.
SZSC is an alloy wire with Sn-Zn-Sb-Cu,Aproper amount of improvement elements are added to the tin zinc based alloy to inhibit oxidability. Various studies show that the current speed can be reduced.improve the performance of the alloy and wire is wildly applied with high cost performance.
In past 20 years,Our company make a various of SZSC prodouct,SZSC-1,SZSC-2A,SZSC-2B,SZSC-3, SZSC-4, SZSC-5, SZSC-6, SZSC-8.Can help the customer reduce the cost one by one.It is one of the best underlayer metal spraying materials for pure aluminum film.
Our product conforms to the requirements of EU ROHS,REACH and relevant domestic laws and regulations for environmental protection.
Copper Alloy Wire,Beryllium Copper Wire,Brass Wire,Tinned Copper Wire
Shaoxing Tianlong Tin Materials Co.,Ltd. , https://www.tianlongspray.com