Microsoft Lee:. NET implemented network will no longer be a bubble

Network | Microsoft Microsoft Lee:. NET implemented network will no longer be a bubble
(2001.12.30) From: Sadie




A few days ago, in China Taiwan media targeted. NET architecture, the former dean of the Chinese Academy of Microsoft, the current Microsoft Headquarters Natural Interactive Service (Natural Interactive Services Divison) Vice president of Kai-Fu Lee, said Microsoft. NET has four elements, is expected to be effective after 34, by then, the Internet will no longer be a bubble, because the future level of network services and today is not the same, absolutely can attract users to pay the cost. In four years, he reckons, the web will be a more high peak.






40-Year-old Kai-Fu Lee, born in Taiwan Province, China, after graduating from primary school to the United States, after obtaining a ph. D. In computer Science from Carnegie Mellon University, has issued the first continuous speech recognition system in the world with Speaker-Independent (speaker-independent). He has served as vice president of Apple Computer company, president of SGI software department Cosmo Software company, and is a Fellow of the United States IEEE. After joining Microsoft in 1998, he was responsible for building Microsoft's China Research Institute, one of the world's four research centers, which was upgraded to Microsoft Research Asia this October.



1.5 ago, Microsoft proposed. NET architecture, Kai-Fu Lee was also transferred to Microsoft headquarters responsible for both technology and products of the natural Interactive service department, currently managing more than 300 engineers, of whom nearly 100 have a doctorate, many are from Microsoft Research turned.



The following is a summary of the interview:



Ask:. NET what is the current state of development?



For:. NET is a step-by-step, very long-term development, there are four things we want to do:



First, the online data (or information) into a structured, not like now rely on a website, or to buy flowers, input "buy" and "flower" and other keywords, we want to make the entire network data into a database, which depends on the XML standard;



Second, the combination of XML into a network service (Web services). Many services cannot be done by a single company, and Windows can be successful by becoming a library that allows everyone to reuse it. Network services will also become a piece of, can be reused.



For example, I hope that when the price of Microsoft is 80, the system can keep me informed, this involves who I am, where I am, how to pass it on to me, what kind of shares Microsoft is, how many prices we have, and so on, each part of which is a problem that can be solved, but these should not be done entirely by stockbrokers, but by the sharing of responsibility by different operators. This reliable soap and other standard solution;



Third, the Network Service personalization, privacy. I hope that when I use the browser "My favorite" is the same on every computer, not a different computer. That is, it should be central to remember what I want, central store my personal data, but also take into account the right of privacy. Now there's my data all over Amazon and on ebay, and the future wants to be the central validation of who I am.



Q: This is the concept of Microsoft's passport, but some have criticized Microsoft for managing everyone's privacy. A: Today's passport is not equal to tomorrow's Passport,passport is only a small part of this concept. We will provide a framework for everyone to federate, and only the operators themselves can manage, but it is the center laboratory certificate, but also provide personalized service.



The next fourth part is to find the most appropriate interface to achieve the first three, which is part of my responsibility. We will use human language and speech to upgrade the current graphics interface mode.



Human language has been developed for thousands of of years, is the most natural way of communication, like a mouse just very low-level control. What we are looking for in the future is not a more beautiful window, a more useful mouse, but a higher intelligent control.



I'd like to have this interface on my PC in three or four years, but that doesn't mean we're not making a product now. We also face some challenges, such as our survey shows users do not like to speak to the computer, how to overcome?



Reduce the fear of the user, one is in the language (with typing, etc. input), one is using voice (input). In language, search is really a very good concept, the user is not disgusted with the search, but now the search is short, slowly can be longer, so "search" can gradually become "instructions."



At present, most of the users are willing to talk on the phone, and not it to say, I repeat, but can use the screen, we are now slowly using the screen phone, you want to check your own transactions today, after voice command, it can list all the records on the screen.



On the phone, you have to enter the "volume of the largest stock" of the text is very difficult, in thousands of stocks with a move down the choice of the way is not possible, or with the easiest to speak. We just have to make a way for the normal phone to be available to the most expensive computers.



Turning speech into an interface, the most important of which will be used on phones, mobile phones and, of course, on computers, takes about three or four years to process.



Q: What do you think of the present. NET the biggest challenge is where? What's good for everyone?



A: Actually, the four points I'm talking about are the main challenges. I hope. NET can solve the problem of network bubble. The reason for the bubble now is that there is no money (free), because it is not good enough to serve. But the future service can jump very big, the user will also be willing to pay money. If there is a system that can be combined with e-mail, voicemail, and phone calls, you can notify me immediately if something urgent happens, I'll pay for it.



That is, many new business models will emerge, and I think it will be four years before the new peak of the network, and the most important thing about the. NET plan is that it can be implemented gradually.



Q: You are a speech recognition expert, is it itself currently in the technical bottlenecks, there is no new methodological improvement? What realm can ai develop?



A: Now the technology can do a lot of things, perhaps now the machine's ability to identify with people 10 times times the difference, but we can set scope to solve, for example, in the stock, there is no "cabbage" and other words. With the current approach, with faster microprocessors, larger memory, and better research, speech recognition is progressing every year. Like Microsoft this year, speech recognition technology is one-fold better than last year, making a mistake by half.



Methodology, the statistical method I did in my doctoral dissertation in 1988 has not changed until now. The ultimate application limit, to achieve human-like thinking, may not be able to use statistics to make. Like people have the ability to common sense, I think in my life should not be able to see on the machine, but the machine does not have to understand the whole thing, can do a lot of things. (Semi)