Last Friday, Microsoft held an adult ceremony for its artificial intelligence robot "Little Ice."
Since its release on May 29, 2014, “Little Ice†has been updated to the fourth generation, and its age has grown from 16 to 18 years old. In the past two years or so, Xiao Bing has accumulated tens of millions of users on platforms such as WeChat, Weibo, and Meitu, and the number of conversations has exceeded 20 billion.
In the process of mutual dialogue, she also appears more and more like a normal person. For example, she can identify the pictures you send to her, can switch between text and voice, can use some kind of "strategy" to bring you negative emotions, and even call you.
"Xiaobing" has become a strategic product of Microsoft in the field of artificial intelligence, and as one of the leaders of this project, Microsoft's global executive vice president Lu Qi also flew from the United States to Beijing. After the press conference, he talked to us about Xiao Bing, Microsoft's artificial intelligence strategy, and how to transform products and businesses over the past few years.
Lu Qi believes that artificial intelligence will eventually be expressed in the core point of "dialogue". When artificial intelligence can naturally talk to human beings and create emotional connections with humans in the movie "She", it will bring huge changes to the whole society. And Microsoft Xiaobing has been on this road for more than two years.
Lu Qi worked for Yahoo for a long time before entering Microsoft. When asked how the giant Microsoft should turn around and avoid the tragedy of Yahoo, he felt that the situation faced by the two companies was not the same. Yahoo's portal has been replaced by search engines and social networks, and Microsoft has taken the original product back to a broader market with cloud computing technology.
The following is an interview record.
About Microsoft Xiaobing
Q: A big feature of Xiao Bing is its emotional computing framework. But human emotions are complex, including positive emotions and negative emotions. From now on, Xiao Bing basically shows positive emotions. Dialogue with people may involve negative expression. Does Xiao Bing learn in this respect?
Lu Qi: According to our artificial intelligence, big data and machine learning technology, as long as there is digital intervention in human emotions, the norms of behavior can basically be learned. Microsoft wants to build artificial intelligence to represent positive energy, to serve humans and help humans. So in this case, we want to establish a dialogue that adopts a strategy that basically guides humans out of his negative or negative environment.
Q: Xiao Bing can already make a call with humans. But the exchange of words between people and machines, real-time speech, real-time understanding of people's context, and voice feedback, is very difficult in this regard?
Lu Qi: In a sense, we have proposed a new standard for the dialogue ability generated by artificial intelligence. How to be in the real-time direction, very intelligent, just like communication between people, it is technically Very big problem.
We must first understand, speech recognition should be very clear, the speed is very fast; second, the ability to interact is very wide, but how to put these technologies, difficult core technology to a certain level, put it together, It is very important to make the entire user experience really like intelligent interaction between people.
Just now a lady went up and tried it. Xiao Bing may not be able to test today that it is a man or a woman who called me today. There are many things to solve.
Q: Why do you want to give Xiao Bing the ability to talk in real time?
Lu Qi: Human beings are different from any kind of place. We invented language, the close exchange between people. Later, after having a phone call, we can exchange them at a distance, that is, real-time exchange, not only human evolution, but also The generation of human knowledge has played a very important role in the entire evolution of life, and artificial intelligence will become the benchmark bridge for new people to communicate with each other.
In fact, it is quite exciting. Although we have not completely solved the technical problem, we have at least defined this as a new milestone. In the future, whether it is a robot or not, it is a natural form of communication, which can be continuously and in real time. Constantly, without reservation.
Q: How far are we from this ideal state of natural communication?
Lu Qi: According to our current deep learning technology, it is only a matter of time. Input the data and continue to train. Before 2009, the error rate is close to 30%. In the past few years, it has dropped from 30% to 8%. It can be predicted that in another year or two, it will be about 1.5%. This is also ours. Technically solvable.
Our voice team is a global team. Xiao Bing is in the pioneering field. My voice team in the United States said that they must continue to improve the priority of Chinese. Xiao Bing is technically the world leader in user experience. . At the time of the demonstration today, we were still nervous. The error rate was around 8%. Once we were not sure, the next interaction was completely inaccurate.
Q: When developing the Japanese version of Xiao Bing, is it based on the Chinese translation platform, or is it a local language developed from scratch?
Lu Qi: From scratch, the cultural and local environment is completely different. At the beginning, we had discussions with the project leader. Is this product only a Chinese game? We say at least try to see which language is close to us, which is Japan. We will know after the test, this product is definitely global. But each country must start from scratch, because each country's language and culture are different, and you can't move.
Q: Nowadays, artificial intelligence is actually using deep learning methods and using many existing samples. How did Microsoft produce such a sample? Give a picture artificially, or analyze his emotions in a paragraph, and train Xiao Bing?
Lu Qi: From an algorithmic point of view, there are several big ways in machine learning from the whole history. One is that you have samples, and there are very big restrictions on artificial intelligence. AlphaGo's biggest innovation is the use of enhanced learning, which can be self-learning, but it also has its limitations. The limitation is how much environment you explore. If the exploration environment is not big, you learn to learn and learn to be closed. system.
Our company's team has developed and migrated, and I have done a lot of personal work. The knowledge between people can be transferred, and migration must be done. Finally, I have to solve some of the problems. I don’t have any knowledge of people, and no one has taught me. I think about it myself. Therefore, human knowledge is the three parts to sum up it and enhance learning. This is the three-part learning method. We have done a lot of research and development in this aspect of Microsoft.
Q: Xiao Bing's ability in all aspects, the grasp of emotions has greatly improved compared with last year. Has Microsoft ever thought about putting commercialization on the agenda?
Lu Qi: Xiao Bing's profit margin is very large. We have the opportunity to create a new generation of profitable forms. Like search advertising, I think search advertising may be the best representative of profit.
Why the search advertising business is very effective, and the user experience can be maintained at a high level, there are several levels, the first, the information structure. Search ads and natural search results are exactly the same, all for keywords. Second, the user experience is also 100% consistent. Third, the effectiveness of search advertising can be 100% tested with accurate methods. Fourth, the search advertising profit system is to establish a market mechanism, and each advertiser must constantly improve the quality of advertising.
Products like Xiao Bing, its strategic significance, the breadth of the entire ecosystem, and the requirements of the team at this stage, basically focus on the user experience and the health of the ecosystem. If you make profits too early, the user experience will be damaged, and the entire ecosystem will be subject to some constraints. We feel that this is not worthwhile. In this regard, we must make the foundation work hard.
Q: You just mentioned the user experience. There are so many things you can do to make a user experience, and it is a very complicated system engineering. Does Microsoft have a set of standards or methods that it has developed internally to turn it into a plan? Sexual things?
Lu Qi: The core of our entire industrial innovation is the combination of three circles, the first circle is human needs; the second circle is technically feasible; the third circle is whether commercial can establish an effective business, All three must be established. The demand of people is the most difficult to predict, and the most difficult to predict is the future.
At present, American technology companies have the best scores and the strongest development potential. The first is Facebook and the second is Apple. These two have in common. They are product development. Technology is a tool for them. Their company The focus is on how simple and accurate the user experience is. In China, I personally feel that Tencent is the strongest player, and Tencent is completely a product oriented. Zhang Xiaolong used to have a PPT to talk about WeChat philosophy. We translated it into English and gave it to my team.
In the new generation, in the mobile and cloud, mobile phones, watches, these digital user experience is no longer a tool, is an extension of the individual, we are very very demanding. I personally feel that the user experience is simple and precise, and it plays a more and more strategic role for the whole company. Technology is always the most important, but the user experience is the core competitiveness of the next generation enterprise.
Microsoft's traditional culture is more about technology, this is a good thing, we still continue to pay attention to technology, but we must improve our culture, constantly insist on the simplicity of the user experience, the world's most complex things to do user experience Simple, it is absolutely not easy to achieve a simple user experience.
Q: There are a lot of technical support behind Xiao Bing, including the cloud computing platform. The cloud platform is in India, including cooperation with medical and agricultural industries. Will Microsoft cooperate with the Chinese government in this regard?
Lu Qi: Microsoft has invested in cloud computing technology as its core. If there is no cloud technology, large-scale data processing, including voice processing or machine learning, can't push forward completely. Just now you talked about doing medical and agriculture in India. We are very, very confident in China, with the Chinese government, and with Chinese state-owned enterprises, to establish a win-win relationship.
I personally don't know the specific case, but I totally believe that our company is very willing. I personally participate in many medical-related and agricultural-related projects in the United States. I find partners in the United States. Cloud computing, artificial intelligence, and big data technologies really advance the society. It is not only commercial, but also available to the whole society. This direction advances.
Q: What is the current amount of R&D investment on Xiao Bing?
Lu Qi: First, the fight is not to see more people, the key is to look at the quality of the team. Second, you must use a local special innovation environment. Third, Microsoft basically emphasizes global sharing.
From the perspective of Xiaobing development, our current team is also very strong, and the number is not necessarily many, but the quality of innovation and the speed of innovation are very fast. The team also made full use of China's unique WeChat and Weibo platforms. At the same time, we make full use of the global technology, voice is shared globally, and many of the big data are shared globally. Many of our US teams have done a lot of basic research to support the team here. I can't say how much money is spent, but it can be said that we have a strong ability to lead the war, how effective our innovative environment is, and how our global brands are used.
About artificial intelligence
Question: From the perspective of computer scientists, the current solution is deep learning. The process of biologists thinking about humans is not very clear. They do not represent real intelligence. I don't know from the perspective of computer scientists, artificial intelligence. What is the direction?
Lu Qi: Our entire industry has been moving forward, but not everyone is 100% with the same definition. Although people are the best representative of intelligent behavior so far, our industry must have a more ambitious idea.
Why can we invent the aircraft? We don't want to repeat a bird, how does the bird fly? After we find the core ability of the bird fly, we build a stronger entity than the bird. The same is true for intelligence. How do people generate intelligent systems and take this rule? At the same time, using this rule to build a more powerful system will have a positive impact on the whole society.
Q: How does Microsoft think about defining true intelligence?
Lu Qi: The definition of intelligence means that we have such a system that can automatically acquire knowledge to represent knowledge and automatically accomplish what he needs to achieve. This is intelligence.
According to this definition, our industry is in this direction, I personally feel that it has gone very healthy, there are several big links to go forward, how do we build a system, it has a so-called broad sense of intelligence.
Question: What is the difference between the narrow sense of intelligence and the broad sense of intelligence?
Lu Qi: Most of the intelligence nowadays is a narrowly defined intelligent system. For example, there used to be a chess machine to defeat the world champion. Although this chess machine can defeat the world champion, but for another game, how simple is it. And it won't go down. What we are going to build now is an intelligent system that can do many, many tasks. Xiao Bing is an example, but there are many other places to do. At Microsoft, we also invested in a lot of machine learning, artificial intelligence research and development, our aim is to build NGI (Next GeneraTIon Internet, the next generation Internet) through Microsoft products for the benefit of mankind.
Q: When do you think this generalized intelligent system will appear?
Lu Qi: I have spent a lot of energy on my own artificial intelligence. I have done a lot of systems, algorithms, and products. I think I am more optimistic. If our entire industry can go forward smoothly, 25 years, even 25 years. No, we have a system that is generalized. It can automatically acquire knowledge and automatically solve many problems in the world. I think the future world is a very, very beautiful world. I personally feel that China is on this world stage. In the past, the status and importance of being able to play will be greater and greater, because of the talent, data, speed of innovation, and space for innovation, I personally feel that there is a unique advantage.
Q: Can you tell us about the latest developments in Microsoft's machine learning and computing capacity expansion?
Lu Qi: Microsoft is very important in the investment of artificial intelligence. We are not only my team, not just the application team. We have great cooperation with Microsoft Research Institute, including several aspects. First, artificial intelligence The calculation must be advanced; second, we must have software tools, you just talked about CNTK, basically the same as Google (Tianbo) Tensor Flow, we use CNTK to implement Microsoft's core products, including like Bing Many intelligent software in search engines and Office365 use CNTK.
We invested in FPGAs on a large scale, and now at Microsoft, all Bing and Office365 and Azure servers, we have plugged in FPGA plug-ins. FPGA can help us to speed up processing. We are now focusing on accelerated processing, deep learning, and we can train on a large scale.
Q: Why is this plugin so important?
Lu Qi: Let's say that a car is driving, or a plane is flying, it must be decided in a short time, so the speed of the model must be very fast, and the model is very large, a lot, FPGA gives us strategically To make a lot of research and innovation benefits, we are now on Azure, accelerated with GPU, FPGA, and will be available on Azure in the future.
About Microsoft Transformation and New Business Development
Q: You have been working at Yahoo for a long time, witnessing the company's rise from decline to decline. At present, Microsoft is undergoing such a transformation, what kind of lessons do you think it can learn from Yahoo!
Lu Qi: Looking back at the history of the IT industry, every 20 years or so, there will be a new era, from PC to the Internet to the cloud, our strategy has been positioned.
Microsoft and Yahoo have a fundamental difference, Yahoo's products, his portal users have found it not very useful, social networking sites, search engines to replace him. And Microsoft's core products are production-efficient software and services, production efficiency software and services due to the cloud, this business will not only go to the sunset, but will grow.
Q: Can you explain to us specifically the "cloud" to bring about this new change?
Lu Qi: The software for producing benefits is in the era of PC. It can already be found in the office. There can be a PC sitting on the office desk, so there are 1.2 billion customers in Microsoft history, all sitting in the office and processing information. Those people.
With mobile phones, with the cloud, there are actually at least 3 billion to 4 billion people using mobile phones and cloud users. These people need a new generation of tools to help them become more productive, once they take Microsoft’s core business. After changing to the cloud, our business will not only shrink down, or become unprofitable, but our space will become even bigger.
Microsoft really has a different status in history. We also see such an opportunity. Although the cloud has a subversive effect on us, it gives us more space. Seeing this situation, no doubt, all The company has moved to the cloud.
Q: What changes have this transfer brought to content product development and corporate culture?
Lu Qi: For example, I took over the Office team for less than three years. When I took over, Office took a break every year. The entire team did not write any programs for 6 months and had plans for 6 months. We are now starting from January of this year, the entire Office is written every month, and has become a service, which is really a big transformation process for the whole team operation culture, development culture, and how to plan.
Our business model is different from before. Office's original business model sells licenses, and now it is sold to people's services, that is, you buy every month, or buy every year, how to sell the way is not the same.
Q: Microsoft's cloud business has changed from $5 million three years ago to $9.5 billion today, most of which was replaced by Office 365. As the person in charge of this team, how can we ensure that the KPI is guaranteed to be fair and reasonable in the process of the new and old, and maintain the stability and morale of the team?
Lu Qi: Office becomes Office365. In fact, there are two levels of parallelism. The first is the existing customers. From the original products to new products, that is, you don’t make much money, just change the way you make money. The new way to make money, the money is still made by Microsoft, and the customer experience is better.
Our KPIs have two major categories - grades and interests, and how many business achievements, this does not mean the future, this represents your past. Equity is another pair of KPIs, that is to say, the number of users, we pay more attention to the quantity, we each look at the number of us, the most powerful one I personally must pay attention to the rights and interests, there are no achievements If you have an interest, your company's future prospects will not be very long. Therefore, the two aspects together are the achievements and rights, and the overall KPI is measured. This is the most important point in the transformation of the Office team.
Q: How do you ensure that customers will use Microsoft products for a long time without turning to competitors?
Lu Qi: Now the cloud service is a real-time, the same is an Office, the same is an e-mail, once in the cloud, how do you use your mailbox, what kind of information you usually send in the mailbox, this is you The data, but we can do real-time analysis, we can build a knowledge map.
Let's say you are a sales team, how do you work with your customers, how do you work with your team. Once we have established such a model, we can do a lot of predictions and a lot of automation.
So after entering the cloud, we also did a test in business. Once Office365, the same customer, we think that in the next ten or twenty years, it may be six hundred percent growth, we can put the level of service Provided very high, mainly relying on artificial intelligence, relying on big data to improve.
Description of Metallic Stainless Steel Braided Sleeving
Stainless Steel braided Sleeve is used as flexible jointing wire of electric appliance, switch appliance electric stove and storage battery etc. such as Cable, CRT Assembly Wire, Television, Computer Monitor, ETC signal pickup instruments. Braided copper wire can be provided rolled flat or in tubular form. Flat braids are generally used for flexible connections, electrical leads and grounding straps. The tubular braids are basically used as shielding.
Features of Metallic Stainless Steel Braided Sleeving
High physical strength with light weight and a long usage life.
2. Good corrosion characteristics, resistance to high temperature, moisture, abrasion and penetration.
3. Good flexibility, absorb effectively vibration, noise, thermal expansion and contraction from pipeline system.
4. Quick alternative for rigid piping in difficult locations, resolve minor deviation because of piping connecting and eliminate the pipeline residual stress, corrects problems of misalignment.
Metal Braided Cable Sleeve,Metallic Braided Sleeving ,Metallic Pet Braided Sleeving,Automotive Metallic Braided Sleeving
Shenzhen Huiyunhai Tech.Co.,Ltd , https://www.hyhbraidedsleeve.com