Shulou(Shulou.com)06/03 Report--

Although the protagonist is still "Internet" at this year's Wuzhen conference, there is no doubt that these Internet companies are swaggering to the pinnacle of new technology. Among them, AI is an important direction. From DeepMind's ability to surprise the world with AlphaGo in one fell swoop, to Facebook, Baidu, Microsoft, JD.com and other Chinese and foreign manufacturers scrambling to set up research institutes, they frequently presented their results at the top meeting. The sufficient data and computing power of high-tech Internet enterprises, coupled with the reserve of talents, make it an important promoter in this round of AI development cycle.

The complete development cycle of a technology usually goes through the process of academic research and development, landing popularization, and breaking through again. Nowadays, the expectation of the whole world for AI is to let it really go deep into the scene and change the efficiency of social and economic development. The focus of the whole society on AI has changed from "how good" to "how to use it". As for the application of AI technology, Dr. Zhou Bowen, vice president of JD.com Group and president of JD.com artificial Intelligence Division and director of JD.com artificial Intelligence Research Institute, put forward a new concept-trustworthy AI at the conference, emphasizing that the future decade of intelligent economy "begins with trust, better than pattern".

Will the word trust be the key fulcrum to help AI applications pry social and economic development?

Solving the crisis of trust is the first step for technology to change the world.

In retrospect, almost every technology will encounter a "crisis of trust" in the process of development. For example, people once thought that photography would steal the soul; when Guang Xu built the first telephone line in China, ordinary people even thought that living near the telephone line would be "bad luck."

Of course, today, with the general improvement in the level of science, technology and education, people's trust in technology has long gone beyond "metaphysics" and began to pay attention to some more practical issues. For example, whether the AI technology performs as well in the real scene as in the laboratory, and whether it can ensure stable and error-free performance in the process of long-term application; whether the application cost and threshold of this technology will be too high, resulting in the unfairness of social and economic development; and whether the technology itself is explainable, and whether the technology supplier is able to bear the relevant responsibility.

In recent years, related discussions have emerged one after another, such as the "people-oriented AI Research Institute" set up by Li Feifei at Stanford. Dr. Zhou Bowen proposed that Fairness (fairness), Robustness (robustness), Value Alignment (value alignment), Reproducibility (replicable), Explainability (interpretable) and Accountability (responsible) are the global consensus on trustworthy AI.

And when people's consideration of a technology begins to develop to the level of trust, it also means that the technology is being applied further and further, from the simple "top ranking" and "attracting Daniel" to win capital attention. development in order to provide reliable technical services to strive for customer choice, as Dr. Zhou Bowen said, "from capital as the core to value as the core."

This change is good news for society as a whole, and technology that has appeared countless times in science fiction has finally begun to enter the scene to change the world. But for the science and technology Internet enterprises, as the driving force behind it, the change of the core driving force of the intelligent economy often means the "change of track" of the development model. Many enterprises should develop from blindly pursuing the high point of technology to considering the comprehensive trustworthy ability.

JD.com who does not need to change tracks

But JD.com is a typical enterprise that does not need to change tracks. Before we go any further, let's take a look at where the power of trustworthy AI comes from.

To put it simply, we can divide the source of this capability into three levels: technology, experience, and concept.

In many cases, the strength of AI technology itself is strong enough, which basically determines the robustness and replicability of AI technology. Take image recognition, which is now widely used as an example, whether the algorithm can identify objects under various angles, clarity and lighting conditions depends on the accumulation of image data. If the amount of data in a certain field is generally scarce, then rely on image enhancement and other technologies to create data. In short, AI technical ability is the base of trust, if the technology itself is not strong, experience and ideas can not play a role. When going deep into the scene, the consideration is the experience of the enterprise in the service. Whether to understand the needs of different industries and make adjustments to the technology, whether they have the ability to explain the technology to industry users, and whether they can balance the costs and benefits of technology application for industry users. These perceptions often rely on experience to accumulate. Finally, of course, there is the idea that if the technology providers themselves do not have the values of technological fairness, the development of AI may really lead to a huge gap between the rich and the poor as many people imagine.

As an enterprise that does not need to change tracks, we can clearly see the accumulation of these three levels of ability in JD.com.

First of all, from the perspective of technical capability, when JD.com officially opened the artificial intelligence technology platform NeuHub last year, JD.com 's own development logic in AI technology also seemed very clear. Led by Dr. Zhou Bowen, former chief scientist of IBM Watson Group, he Xiaodong and Mei Tao were introduced into speech and semantic natural language processing and computer vision respectively, which further strengthened the talent pillar for JD.com AI. Four major R & D centers have been laid out around the world, including Silicon Valley, Beijing, Nanjing and Chengdu, allowing NeuHub to gradually establish its own technological highland. Coupled with the data advantages accumulated by JD.com e-commerce, it is enough to provide reliable technical support for reliable AI.

What can help JD.com accumulate reliable AI ability must be JD.com 's own service experience. Throughout a series of businesses under JD.com Group: e-commerce, finance, logistics, health, insurance, overseas markets and so on, all of them are service-oriented. For the excavation of different scenes, on the one hand, it opens the way for JD.com to directly access industry users and even individual users, so that JD.com can input rich technical capabilities into it; on the other hand, it also lets JD.com know their real situation: what kind of help different industries need, what kind of technical application conditions they have, and how much cost they can bear. Based on this accumulation, JD.com can finalize the three core business directions of intelligent interactive RPA, video Internet of things and intelligent supply chain, with targeted output capabilities for different scenarios.

As for the concept of development, as one of the national new generation of artificial intelligence governance professional committee, Dr. Zhou Bowen emphasized harmony and friendship, fairness and justice, tolerance and sharing, respect for privacy, security and controllability, shared responsibility, open cooperation, agile governance and other eight principles based on the "new generation of artificial intelligence governance principles-the development of responsible artificial intelligence". Use it as a benchmark for measuring all initiatives.

Today, when AI is rapidly moving towards the real world, JD.com is making full preparations for the next decade. The technical capability itself, the way technology is combined with the whole world, and the Vernier calipers that allow principle to measure all these, all embody the standard form of reliable AI.

The pattern and mood of the ladder maker

Where do the technologies that successfully build trust end up? Looking up at the electric lights overhead and touching the high-quality and inexpensive clothes, we can find that those technologies that are influential enough have finally revolutionized by subverting production efficiency and continue to benefit the world a hundred years later. and a series of technological sequences have been derived upward to drive the whole world forward.

From the perspective of scientific and technological Internet enterprises, when the industrial Internet becomes an important target, industry users will be their important service objects. The service providers faced by different industries, such as health care, education, finance and so on, are every ordinary person like you and me. This is another important reason why AI needs to develop in a reliable direction. In addition to entering the real scene and driving the development of social productive forces, the technology that can really be called the industrial revolution should share the dividend with every ordinary person.

In the technical layout of JD.com AI, we can find traces of this trend.

For example, in JD.com 's intelligent supply chain system, it can help small and micro merchants accurately estimate the scale of sales, and even a small vegetable store can keep the ingredients fresh and increase turnover. At present, retail has entered the second half, AI can match an AI assistant for each item, which can be matched to the most suitable users according to the different tonality and characteristics of the goods, so as to reduce the cognitive burden of users. East AI's pharmaceutical knowledge graph based on NLP technology can help doctors and patients use drugs more safely. As for garbage classification, which gives many people a headache, JD.com AI found a breakthrough through image recognition technology, packaging the "garbage recognition algorithm" into garbage classification API and Mini Program to help people solve the thorny problem of garbage classification. With the combination of AI, Internet of things and blockchain, the circulation of agricultural products can also be visualized and traceable, thus alleviating people's concerns about food safety. Intelligent customer service's understanding of Sichuan dialect not only makes users feel more friendly, but also expands its own service group, so that users with less standard Putonghua can also enjoy the convenience brought by AI.

For industry users, AI trust may be a systematic consideration, but for consumers, trust in AI comes from the accumulation of each tiny experience and grows in detail. This trust not only shortens the distance between people and technology, but also shortens the distance between people. People can rely on the freshness of the ingredients in a small store, the safe source of the goods in hand, and the intelligent customer service provided by the merchant. In many cases, the change in efficiency begins with mutual trust. When people no longer question and verify repeatedly, the flow of productivity and goods will also become smoother.

When we talk about AI today, we are actually talking about the industrial revolution and the opportunity for a change in the world order.

For China, this is the first time we have been fully prepared for the advent of the industrial revolution. The AI we want to build is not only a magnificent "academic palace" or a solid "economic fortress", but a ladder that can be interspersed with all buildings and continue to develop upward, so that every ordinary person, every industry and even the whole country can move up to a brighter place.

And the most important thing to create a ladder is to let people maintain trust and take the first step up the ladder. This is the pattern and state of mind of ladder makers like JD.com.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.