How to crack artificial intelligence "difficult to land"?

The current artificial intelligence is like the Internet in the 1990s.

The outbreak of any new technology in history has brought about new business and new economic models beyond imagination. From a recent perspective, in the early days of Internet development in the 1990s, we did not expect the commercial society to be affected and changed on such a large scale. The influence of the Internet began in the media. The New York Times and The Wall Street Journal and other media updated the news through the website, and eventually the entire media industry is facing transformational pressure. Later, consumers gradually bought books and rented video tapes through the Internet. The impact of the Internet on the business community is a process, as is the era of artificial intelligence.

A film lover who cares and often forgets to return the video on time and has to pay a large fine, Reed Hastings, founded Netflix to avoid the high-expected fines of the movie rental model. In 1999, the recently established Netflix launched an online video subscription service. At that time, Blockbuster, the largest film rental chain in the United States, did not expect Netflix, which had less than one-thousandth of its turnover, to become a future industry disruptor. By 2007, Netflix had more than 7.5 million registered users, a compound annual growth. The rate is higher than 50%. In 2010, the American DVD rental giant Blockbuster filed for bankruptcy. In 2011, Borders, the second largest chain bookstore in the United States, filed for bankruptcy. In the early days of Internet development, offline business giants did not expect the impact of new technologies and business models to be so great. They did not regard the new model on the emerging Internet as a potential adversary and attracted attention. After more than a decade, these former giants have disappeared. .

Zhang Yizhen believes that the same thing will happen in the era of artificial intelligence. Even if the company with a large volume is unable to grasp and understand the new trend, it will be subverted in this wave. Conversely, like Amazon, Netflix Because of the new technology, there is an opportunity to subvert different industries.

In which industries will the artificial intelligence flower bloom?

At present, some applications related to computer vision, such as in the security field, have produced relatively large changes. The previous security industry used the idea of ​​tracing responsibility afterwards, looking back at the historical records by looking up surveillance videos. For example, Beijing Airport is said to have more than 20,000 cameras, which is impossible to rely on human real-time monitoring to ensure security . At present, artificial intelligence can prevent and prevent some dangerous situations from happening in real time.

In the future, artificial intelligence will have a great impact on industries such as finance, healthcare, education, manufacturing, retail, transportation (autopilot), and logistics. Zhang Yizhen believes that in the stage of artificial intelligence algorithms relying on big data training, the industries with large value will develop well--if the value of artificial intelligence generated by an industry is large enough, some people will be willing to pay for the data. Most of the artificial intelligence now relies on a large amount of data to learn. The premise of developing an artificial intelligence-related application in an industry is to obtain data that is related to the industry and the field, and is labeled and collated.

Microsoft hopes to work with a few leading companies in the financial, medical, manufacturing and other industries to clarify the problem and give an effective solution to the problem. After that, it is possible to modularize the solution and let more Partner utilization. At present, Microsoft Research Asia has established an organization “Innovation Meeting” to promote the development of industry applications through open innovation, to strengthen cooperation with leading companies in various industries.

Every wave of technology will dramatically increase social productivity. Compared with 20 years ago, the Internet has greatly increased social productivity. The same is true for artificial intelligence, but the artificial intelligence landing requires deep cooperation between technology and industry companies. To achieve the improvement of the whole society's productivity, one or two enterprises can't do it. It requires many forward-looking leaders to cooperate.

Take the artificial intelligence application in the financial field as an example. In terms of fund management and auxiliary stock analysis, technology companies and financial companies have their own strengths. Some fund companies also have teams of data scientists, some do more traditional data mining, and are not very familiar with artificial intelligence technologies such as deep learning. Based on decades of experience, fund companies are better at judging whether a stock is worth investing and which market information has reference value. Artificial intelligence technology companies do not have the knowledge accumulated in this area, but have advanced artificial intelligence technology reserves. Through the cooperation between business companies and technology companies, fund managers can better analyze the market with the help of artificial intelligence. For example, some listed companies said that due to the late Spring Festival this year, which affected the sales of this quarter, analysts need to make an analysis and historical comparison on the impact of the Spring Festival on sales, and analyze whether this is the company's excuse or the real situation. Every listed company will have a quarterly report, ranging from a few pages to dozens of pages. To do a very detailed analysis, in addition to looking at the present, we must compare the data of last year, the previous year, and even the previous year, so detailed analysis. It is impossible to rely on people alone. An analyst has to analyze dozens of companies. It is impossible to look at each quarterly report very carefully. This aspect can be assisted by artificial intelligence.

The lack of artificial intelligence technology companies to enter a certain vertical field is the data and knowledge of related fields. If you work with a hospital, because the technology company does not have an experienced doctor, it is impossible to judge whether the medical image data is correct. Due to the lack of relevant professional knowledge and experience, it is impossible to judge whether it is due to mislabeling or because the image is not clear enough. When technology companies interact with vertical industries, they need to understand the industry. Artificial intelligence is not superhuman wisdom. It is impossible to provide a database to the machine to get the desired results. This is one of the challenges facing the current cooperation. . Since the data needs to be labeled, the premise of artificial intelligence implementation at the current stage is based on a large number of artificially labeled data. For example, millions of photos in ImageNet, the world's largest image recognition database established by Stanford University, are also used by many people after they have spent a lot of time marking the machine.

Different development paths of artificial intelligence applications between China and the United States

The development of artificial intelligence in different countries is related to the characteristics of local industrial development, depending on the combination of technology and local industry. Take the financial industry as an example. There are two major differences between China and the United States. First, in terms of technology applications, the US financial market is highly competitive, and many banks have long been accustomed to competing through technological means. 10% of employees in a financial company are IT and technical employees, and in China, the ratio is about 3%-4%. In the United States, the application of artificial intelligence in finance is relatively advanced. Many hedge funds manage funds through machine learning, data mining, and quantitative funds through procedures. Compared with the United States, China is relatively early. On the other hand, there are certain differences in the regulatory regulations in the financial sector between the two countries. In the United States, there are not too many regulatory restrictions to manage funds through development procedures. As long as they dare to take risks and are responsible for their own profits and losses, China is relatively cautious overall.

In other application areas, China and the United States also have their own characteristics. For both China and the United States, the demographic dividend is disappearing, but the application of artificial intelligence in the two countries is likely to be developed first in their more developed industries. In the United States, the service industry is relatively developed. At present, the application of artificial intelligence is more to consider the application of robots from the perspective of the service industry. For example, taking care of the elderly in hospitals and nursing homes. In China, manufacturing transformation has become a trend, and the repetitive work of manufacturing workers is no longer attractive to young people. In Shenzhen, many manufacturing companies are not happy with employees. Not only in China, but also in some developing countries such as Vietnam, there will be similar problems. In this case, the manufacturing industry will rely more on artificial intelligence and other technical means. In the future, China will mature these technologies first. It is possible to apply technology to other countries.

How to judge whether the technological innovation in the field of artificial intelligence can land?

How to judge whether the artificial intelligence technology development currently underway has the possibility of landing? For example, can the speech recognition technology be applied to the online translation of the conference? Zhang Yizhen said that Microsoft Asia Research Institute uses BTX when commercializing the technology innovation. (Business Business, Technology Technology, Experience User Experience) judgment principles. The first step is to judge whether the technology is mature. Can this scene be realized? Automatic recording and automatic conversion into text during the meeting. There are many factors involved in realizing the application of the technology in the real scene. It is possible that the Chinese content of the speech is mixed with English. It is possible that the speaker is far from the microphone and the voice is unclear. So judge whether the technology itself is mature? If the product is to be made, can the technology itself be achieved? Second, if the technology is achieved, what is the user experience? Will the user use it? If the technology and user experience of the product itself are not The question is to consider whether there is a way to generate a certain amount of income, so that the product can be maintained and continuously improved. For example, the example of real-time translation of the conference, if the service is priced at $3,000 per hour, it is difficult to sell. But if the price is $3 per hour, it is very likely. The first is technology; the second is the experience in the scene, whether to make what the user can use; the third is the operational business model, is it possible to enable the service at a cost that the user can accept.

Over the past year or so, many reports on artificial intelligence have led to some misunderstandings among the public. Can artificial intelligence become the first place in the world of chess to mean that artificial intelligence is smarter than anyone? Many people can learn to drive, but it is much more difficult to drive a computer safely than to win a game. Go is a finite variable, and driving involves more judgment. If someone on the side of the road beckons at you, is there a policeman who stopped because of an accident? Or someone wants to hitchhike? Or someone is broken, please help me? To understand this scene, you need more variables. Is this person wearing a uniform? Is there a car that breaks down? These variables are ever-changing. The computer can't understand what this person's expression looks like. What does the expression mean? The so-called "common sense" of human beings is very difficult for the computer. .

The top leaders in traditional industries, especially enterprises, need to have a more objective understanding of artificial intelligence. As far as possible, they should contact and understand what artificial intelligence can do, what can not be done, and should not have high expectations. After all, some technologies have not yet reached maturity. There are a lot of artificial intelligence online courses and books, and there are many free online platforms to encourage everyone to try. This threshold is getting lower and lower. In the transformation of traditional enterprises, the first is to understand the technology, the second is to understand their own industry, and to think about what problems are solved by artificial intelligence is most valuable to their own enterprises. In order to attract eye-catching projects, it is better to be a key project that can be effective for the company. The success of the transformation will be much greater.

This wave of artificial intelligence is similar to the early days of the Internet. Regardless of the size of the enterprise, it is not too early to embrace AI at any time. Enterprises will eventually increase production efficiency due to the application of artificial intelligence. Large enterprises have their own IT departments and resources. They have the conditions (human and related resources) to study AI. You can choose to do it yourself or seek external help. Small businesses may need to find someone to do it. The thinking mode of large and small enterprises is the same as the basic logic, but the specific operation modes are different.

Regarding whether a traditional enterprise needs its own technical team, it should be combined with the company's own situation. How difficult is the current project to be done, and whether it needs expert help. Dr. Zhang Yizhen has seen many traditional corporate executives who are generally interested in artificial intelligence, but there is still much to be done to understand artificial intelligence. Enterprise transformation AI is like a person's health problem. Everyone should have basic common sense about their health. See if they are difficult to solve or find an expert. Enterprises have this premise, understand their own problems, understand how technology is applied, and further analyze the difficulty and risk of judging things. For example, there are a lot of discussions about chat bots. Microsoft has chat bots like Xiao Bing, so many companies have high interest and hope to use chat bots to enhance interaction with customers. Then you need to consider what the content of the chat is. What is the cost of the mistake? If the hospital needs to interact with the patient and instruct the patient how to take the medicine, the cost of the mistake is too high and it is not recommended to develop through the internal team.

On the eve of the take-off of artificial intelligence technology applications, the value of such discussions is that traditional companies, regardless of size, need to think about how to avoid becoming the next Blockbuster or Borders, and whether they can seize the opportunities provided by the technology wave and realize the transformation of business models. After all, when the users brought by artificial intelligence technology reach a certain concentration and the new business scale takes off, it is too late to make such an investment.