Decryption of Microsoft Mini-ice: Large data for artificial intelligence

July 2, the Microsoft (Asia) Internet Engineering Institute announced that the second generation of intelligent chat robot "Microsoft Ice" officially on the line, users can log on to Microsoft Ice official website for "adoption" operation.

So what is Microsoft's small ice product? First, Microsoft Ice is a smart chat robot. In addition to intelligent chat, but also a reminder, weather, constellation, traffic guide, restaurant reviews and other practical functions. Second, Microsoft Small ice integrates Microsoft's technology accumulation in large data, natural semantic analysis, machine learning and deep neural network, and has very powerful machine learning ability. So, Microsoft small ice is an artificial intelligent robot.

"Artificial intelligence needs to be supported by big data."

Artificial intelligence has three main branches: first, rule-based artificial intelligence; second, there is no rule, the computer reads a large amount of data, through the data statistics, probability analysis and other methods of intelligent processing of artificial intelligence; third, a deep learning based on neural networks.

Based on the rules of artificial intelligence, that is, in the computer according to the prescribed grammatical structure input rules, and these rules for intelligent processing, lack of flexibility, not practical. Therefore, the actual mainstream of artificial intelligence is the latter two. Both then read a lot of data through the computer to improve the accuracy of the AI itself. Today, after a lot of data generation, there are low-cost storage to store it, with high-speed CPU to deal with it, so the artificial intelligence after two branches of theory and practice. Thus, artificial intelligence can make close to human processing or judgment, improve accuracy. At the same time, the use of artificial intelligence services as high value-added services to attract more users of the main factors, and the increasing number of users will produce more data, so that artificial intelligence further optimization.

Large data is divided into "structured data" and "unstructured data". "Structured data" refers specifically to data that can be managed as a database, such as customer information, business data, sales data, inventory data, etc., stored in a common database. In contrast, "unstructured data" refers to data that is not stored in a database, including e-mail, text files, images, videos, and so on. Currently, unstructured data is proliferating, and about 80% of enterprise data is unstructured. With the advent of social media, unstructured data has been greeted with a burst of growth. Complex, massive amounts of data are often referred to as large data.

However, the analysis of these large data is not simple. Text mining needs "natural language processing" technology, image and video parsing needs "image parsing technology". Today, speech recognition technology is also indispensable. These are the techniques studied in the field of artificial intelligence in the traditional sense.

Microsoft's small ice is also entirely because of the support of large data to achieve artificial intelligence. According to online media reports, Microsoft's small ice set up nearly 700 million internet users in China over the years accumulated all the public chat records, and refining out of 15 million corpus, based on data mining and intelligent search, through understanding the context and semantics of dialogue to achieve the human machine question and answer the natural interaction, or even the equivalent of 16 years old girl's IQ.

"The combination of big data and artificial intelligence will unleash greater energy."

In addition to Microsoft, Google and Apple are also working on the development of products that combine large data with artificial intelligence. Google to provide optimized search engine services, the background of artificial intelligence with the user's continuous evolution, the use of more users, the search engine will be optimized, optimized, the user naturally more. In addition to search engines, Google also through Gmail, Google Docs, etc. to obtain a large number of "unstructured data." As a result, Google's brain becomes smarter. In addition, Google has developed a "semantic search" of the evolutionary system. Apple's speech recognition technology Siri is also based on the latest AI theory (deep learning). In turn, the evolution of modern artificial intelligence requires not only theoretical research but also a large amount of data as a support.

At the beginning of 2014, Google joined Audi, GM, Honda, Hyundai and Nvidia to set up a new consortium-the open Automotive Alliance-and Apple began dabbling in automotive intelligence in June 2013, launched the "IOS in the car" technology, through the Siri voice operation can achieve navigation, call, music playback and other services. Recently, Google announced that it will be in 2017 to the market to invest in self-driving cars, as of August 2013, has completed a 480,000-kilometer test drive. The large data of 480,000 km test drive has become the driving experience data, which provides the decision analysis basis for the automatic driving of artificial intelligence.

In the future, all industrial areas will be intelligent through artificial intelligence and artificial intelligence robots. Artificial intelligence will become an important source of high value-added, play a huge role, is likely to exceed the world has caused great changes in the "Internet revolution", the impact will be enormous, will also be spread across all levels of society. Google and Apple have brought a deep sense of crisis to traditional industries such as automobiles. And their subversive innovation, in fact, from the combination of large data and artificial intelligence. Perhaps this is well worth our thinking. Will the same phenomenon and the same subversion happen in more traditional industries?

June 12 Afternoon, the author has been in the second meeting of the CPPCC to observe the 13th session of the CPPCC Consultative Forum. At that time, JIU, vice president of Central Raiming, made a proposal to "implement large data development strategy as soon as possible". The author hopes that artificial intelligence can become one of the foothold of large data development strategy.