If its a scene that even a science fiction movie can describe, its bound to come true in the future. Its just when and to what extent.
On July 22, surging journalists had a 40 minute exclusive in-depth conversation with Shen Xiangyang, former executive vice president of Microsoft and chairman of Xiaobing company. When it comes to the possibility of artificial intelligence in the future, Shen Xiangyang made the above remarks.
Recently, Microsoft announced that it would split the AI small ice business into independent companies, and appointed Dr. Shen Xiangyang as the chairman of the new company. Not long ago, Xiao Bing also received an honorary diploma from the music Engineering Department of Shanghai Conservatory of music, which indicates that AI has demonstrated the same musical creativity as human beings in some aspects.
In the future, the number of AI interactive agents will far exceed the number of human beings. Shen Xiangyang said, the biggest interaction in the future is not the human-computer interaction now, but the interaction between human and artificial intelligence. This will be the biggest opportunity for AI in the future and the ubiquitous entrance that Xiaobing is looking for.
According to Shen Xiangyang, most of the artificial intelligence that can be achieved today is still focused on the work in a special field with large amount of data and computation. But in fact, the real intelligence is still far away from us, such as the problems of common sense and reasoning.
Talking about the future development of science and technology in human society, Shen Xiangyang said that his three most promising directions are artificial intelligence, augmented reality and quantum computing. At the same time, Shen Xiangyang also thinks that the next revolutionary hardware may be smart glasses. If smart computers, smart phones and smart glasses are combined, it will certainly be revolutionary, but it is very difficult to do, Shen said
Shen Xiangyang was once the highest level Chinese executive in Microsoft. He was fully responsible for Microsofts global AI strategy and AI product business groups, including Bing and Cortana product divisions. He also had jurisdiction over Microsoft Research Institute, which was responsible for integrating scientific research with Microsoft product engineering. Shen Xiangyang is an IEEE fellow and acmfellow. He was elected a member of the National Academy of Engineering in 2017. He was elected to the Royal Academy of Engineering (UK) in 2018.
The following is the transcript of the conversation between the surging journalist and Dr. Shen Xiangyang (edited by the editor)
Shen Xiangyang, former executive vice president of Microsoft and chairman of Xiaobing company
Xiao Bing has released so many generations, and his skills in the field of music are relatively mature, which can be regarded as traditional artistic ability. On June 29, she got the diploma of Shanghai Conservatory of music. What new technological breakthroughs has Xiaobing made?
Usually we input a piece of text or a picture to inspire Xiaobings creative inspiration. Then Xiaobing can create a unique music track. She can also automatically complete the orchestration, composition and lyrics creation according to the style and rhythm, and the speed is very fast. Now a three minute song takes only two minutes to write. In fact, we can continue to optimize the speed. In terms of style, Xiaobing has mastered the creation of pop music, folk songs and ancient customs.
What kind of role does artificial intelligence technology play in Xiaobings music creation? What are the advantages of Xiaobings AI technology compared with other AI technologies?
As for the creation and generation of music, we have been thinking about this issue since the advent of computers 50 or 60 years ago. At the beginning, of course, the traditional method is relatively simple. Through splicing and statistical probability, the music that has already existed can be divided into sections and then connected. This practice is relatively poor in innovation and creativity. In fact, the real breakthrough is still in recent years. The development of deep learning and deep neural network makes the learning of artificial intelligence possible. In the past few years, as long as they are effective deep neural networks, we will try to make them generate music.
In fact, which neural network to use is not the most concerned problem for Xiaobing. In our opinion, the most important reason for the good quality of Xiaobing is that,
We have taken a deep neural network, deep learning and music knowledge organic combination of the road.
Because if you only use deep learning, it may be good to do it, but there will always be a little specious feeling,
Music knowledge must be added in order to really do a good job in music.
In composing and arranging music, the most difficult part is evaluation
u3002 Even among professional musicians, peoples views on music are not always consistent. Then we will study with the teachers of the Conservatory of music. This time, we also learned a lot in the process of Xiaobings graduation, including the evaluation of composition and composition, including the integrity of melody, singing, etc. we have made great efforts in all aspects of Xiaobings music creation, so it is not just a simple problem of deep learning algorithm. Of course, we are still improving the algorithm.
Xiaobing has been focusing on the creation of EQ and Humanities and arts, so we will also pay close attention to the commercialization of Xiaobing. Last year, we have seen that in the field of pattern design, Xiaobing has begun to design patterns for garment enterprises. How will she create value by integrating music with the industry?
In the field of artificial intelligence creation, we think that there are great opportunities for industrialization, including poetry, painting, music and singing.
In fact, behind each kind of creativity, Xiaobing can correspond to a production line, and Xiaobing has completed all kinds of creativity modules necessary for production of all production lines
After graduating from the Conservatory of music, we have basically completed the layout in the fields of artificial intelligence text, sound and vision. Next, we will expand these areas to do more content, and even do some cross domain content.
But we can see that one of the areas that have been implemented is education. During the period of Shangyin study, Xiaobing participated in a very remarkable project called when traditional culture meets artificial intelligence as an intangible cultural heritage Music Ambassador. This project is to help the children in these intangible cultural heritage related areas complete the creation of their first songs in their lives. It is very meaningful and moving. Many children have participated in this project. Most people have a little fear before they have done creation. So with the help of Xiaobing, the children in this project can quickly complete the rest of the work as long as they create the main melody according to their hometown music, so that the children can have the first music work in their life to show the style of their hometown family. We think there are many opportunities in the field of music education. Once this capability is developed, not only Xiaobing can do it, but other developers can also have more imagination.
Now, in fact, the competition of voice related track is very fierce. All technology companies say that they are using technology to empower enterprises. What advantages does Xiaobing have in enabling enterprises?
Many companies are doing this work. I think its very good. It shows that we all see such a track and opportunity. Xiaobings characteristics and advantages, the first I think is the technology itself. Because we have been at Microsoft for so many years, we still have great advantages in natural speech processing, computer voice, computer vision and knowledge mapping.
We really didnt take the road of hardware
u3002 In fact, the way to hardware is to subsidize and invest money, which takes quite a long time, but in the final analysis, the key is the experience of voice interaction and what you can do with voice interaction. Xiaobing has been on the road of EQ + IQ from the beginning. To be such an AI assistant, it is very important to have a link between AI assistant and human beings, so that the interaction between AI and human can be better.
In fact, Xiaobing is separated independently. We feel very confident that in the future, the biggest market in the future will be the interaction between human beings and artificial intelligence. There will be a lot of artificial intelligence in the future. Its not only that you have a Siri on your iPhone, on your mobile phone, on your speaker, almost every application will have an AI assistant to push,
In the future, the number of AI interactive agents will far exceed the number of human beings
u3002 The biggest interaction in the future may not be the current human-computer interaction, including the interaction between people and mobile phones, the interaction between people and PC, even the interaction between people and speakers. These are small things.
From the perspective of voice, you must have a microphone to receive voice. If there is input, artificial intelligence can appear in any place with voice in the future, not just in the place with speaker. I used to like the Siri team very much. I have communicated with several of their founders for many years and learned a lot from them. As a matter of fact, I always appreciate their idea at that time that wherever there is a voice, it is the entrance. So the problem you have to think about is not just to make a speaker or just to make a hardware. In fact, it is very natural to have Siri on the mobile phone. It may be a super assistant and very powerful, but this does not mean that human beings only need to interact with an AI. The mobile phone can have the AI in the mobile phone, and the car can have the AI in the car. This is another kind of interaction. Speakers have proved to be an entrance. Our expectation and judgment for the future is that artificial intelligence in the future, no matter what form, will appear more and more. There may be dozens of AI in a mobile phone, PC, speaker and other places. So from the present human-computer interaction to the interaction between human and artificial intelligence, this is the future we see.
Xiaobing has a lot of abilities in the field of art and humanities, and pays more attention to Eq. I remember when Google was doing alphago, the way alphago played chess also inspired many human chess players. Xiaobing pays attention to creation and EQ Development, will it also bring some new inspiration to human beings?
Yes, Im sure it will. Alpha dog is a great thing. He can find out some ways to play chess and play some chess paths that human beings have never played. But relatively speaking, alpha dogs only focus on one thing: go, which is a relatively visible and totally closed game environment. Xiaobings artificial intelligence creation is still different. The biggest difference from playing chess is that it is relatively clear whether the winner or loser is a winner or loser. However, for example, in music creation, it is not easy for us amateurs, such as professionals and music professors, to tell them exactly how you evaluate the music. After communicating with professionals, we tried to divide music evaluation into several categories, such as what I just mentioned is fluency and what is called singing. These are not clearly defined so far. So in this, I think it provides some innovative inspiration, that is, we have more imagination space.
In fact, I very much hope that we can make some remarkable work in the evaluation system and evaluation standards
This is not limited to music, painting, poetry and other fields we have done before. If we still think of ourselves as a pure researcher, we cant do it.
Now that Xiaobing has been able to replace part of human work, is it possible for artificial intelligence in the future to develop into a Robot Companion like that in movies?
I think, if youve seen the movie her, its already happening.
I have a romantic view of technology. I always feel that there is no technology that cant be done in the world, only technology that cant be thought of. What man can imagine, historically, has been realized.
Its just that some things have been imagined for two thousand years before they are realized. For example, Changes flying to the moon, Qianli eye and shunfenger have all been realized today. I think that if all the scenes described in science fiction movies can be realized, it is nothing more than when and to what extent.
Artificial intelligence has begun to replace some peoples jobs, white-collar workers in the office and blue-collar workers in the production line. For example, the most popular RPA (robot process automation) recently is to sort out and integrate the input and output data of general companies. It is very boring work. We can use RPA to replace human employees. On the contrary, these jobs created by artificial intelligence will not immediately replace anyone. Xiao Bing has published poetry collections and painted paintings, but I dont feel that any poet or painter feels threatened. And RPA such work, is a substantial part of the companys employees are affected.
Like robot companion, I think that since Siri came out, we have a very clear understanding of this matter. Once a very famous article in the New York Times was written by a mother. This article is called thankyou, Siri. It tells us that her child has some mental retardation, but communicating with Siri makes her life more satisfying. This article points out a direction for us, that is, artificial intelligence will definitely occupy an increasingly important position in human life, which is also the thinking of our team. In the future, the interaction between human and artificial intelligence may be one of the most important things.
You were responsible for Microsofts global AI strategy and one of the worlds top Chinese scientists. How much impact do you think AI will have on human society in the future?
I have been at Microsoft for many years, and I am very lucky to lead Microsoft Research Institute and Microsoft artificial intelligence product line for many years,
There are three directions that I am very excited about. One is artificial intelligence, one is augmented reality, the other is quantum computing.
Among them, artificial intelligence has had a huge impact on human society.
Today, almost every level of technology in computer science, from chips to algorithms to software to interaction, has actually been subverted. For example, Intels nvia chip market value has surpassed Intel nvia. One of the important reasons is that NVIDIA seized the opportunity. Today, a lot of computation of chip is based on artificial intelligence. Including the Cambrian listing in China a few days ago, we are also full of expectations.
Artificial intelligence has a great impact on human beings, mainly because of the use of data. As long as you can get through all the isolated islands of data, the impact of artificial intelligence on human beings will continue to increase.
Now what most people see is the impact of artificial intelligence on peoples lives,
Next, what I personally think is of great significance to enterprises
u3002 Like the RPA mentioned just now, RPA is moving forward. RPA 2.0 and RPA 3.0 are bound to happen. As an artificial intelligence practitioner, I would also like to remind you that most of the artificial intelligence that can be achieved today is still the work with large amount of data and calculation, focusing on a certain field. For example, alpha dogs can do very well in chess. But in fact, real intelligence is far from us. From the point of view of general artificial intelligence, we have no clear direction at all,
We have not yet achieved real intelligence, such as the problems of common sense and reasoning.
In the next 5-10 years, including Xiaobing, there are many meaningful things to do. We hope Xiaobing can develop faster and better in general artificial intelligence.
I remember that there was a debate between Elon Musk and Mark Zuckerberg, that is, whether AI is beneficial or harmful to the future of mankind. They have different views of optimism and pessimism. What do you think of this problem?
I think so. First of all, these people are very famous people. They are too serious about what they say, especially musk.
I think people have a general fear of the development of artificial intelligence in the future. Fear is due to peoples lack of understanding of artificial intelligence. More importantly, the development of artificial intelligence is a very special thing in human history. The great thing about human development over the years lies in the constant invention and creation of new tools to improve our lives. But before artificial intelligence happened, almost all the tools were extending our physical ability, and artificial intelligence was extending our brain power. Unfortunately, brain science is still very early. For example, in brain science, we cant do a lot of repeated experiments today to really understand whats going on. So when artificial intelligence comes into being, its normal for us to have a sense of fear and wonder whether human beings will be destroyed in a few years.
This is very normal. For example, after the invention of nuclear weapons, it is likely that we will end up with more than a few more nuclear bombs in the world, or some comet will come and hit the earth one day. This possibility will always exist.
The fear of artificial intelligence is very normal, so now many people have been discussing how to standardize it and make some restrictions on the development of artificial intelligence. Including myself, I also agree to discuss and standardize the ethics and Prejudice caused by artificial intelligence
u3002 In the research and development of Xiaobing, I have always stressed that there should be a sense of awe when making Xiaobing. After all, it is connected with peoples emotions and emotions. Whenever we make such a product, we must have a sense of awe.
After the Internet and smart phones, human society has been waiting for the next revolutionary technological change. Do you think this revolutionary change has appeared? If not, what direction will it be?
I think that although the invention of the Internet is not so great in technology, in human history, its impact on human society may be ranked in the top three. Maybe the first one is to make fire out of wood. There can be another one in the middle. Then there is the Internet. It can be ranked at least in the top five. The Internet has connected the whole human being. It is a very, very amazing thing. Since then, human beings have such a feedback mechanism, so it is very difficult to compare anything with the Internet.
From the perspective of interaction, especially from the point of view of hardware, many people are very keen on the issue of smart cars. I also agree that electric vehicles, smart cars and unmanned driving are definitely the future, which is undoubtedly very exciting. but
My favorite hardware, the next revolutionary thing, may be smart glasses, but this is very difficult to do.
For example, if I wear glasses myself, I will know that as long as the glasses weigh 5 grams, people can feel them. So you have to make them very small, which also involves a lot of optical things, which are not easy to do. Im optimistic about the direction of Ar + AI. I think the interaction between people and glasses is very natural. At least its more natural than wearing a watch. You already have a mobile phone and you need to wear a watch. Of course, the watch has its reasons, but relatively speaking, the delta is not that big.
In fact, if a smart computer, a smart phone, and smart glasses are combined, it will certainly be revolutionary. So Microsoft does hololens. I like it very much. But in terms of experience, its still a little bit early for it to become a consumer electronics product. Therefore, products like hololens are still playing in the enterprise market. In terms of computer science, I believe that quantum computing will bring us great surprise.
You announced your retirement from Microsoft last year, and now you come to Xiaobing as chairman of the board. What kind of mood is it to come to Xiaobing?
Ha ha ha, Im glad to be able to do something with you. In fact, I have been away from Microsoft for a short time. I announced my retirement in November and officially left Microsoft on March 1. Now I have the opportunity to work with a group of old and new colleagues to accelerate the development of Xiaobing in the local market, including commercialization. I think it is a very good opportunity. The development of Xiaobing product line in the past five or six years is very good, and it is the star product of Microsoft artificial intelligence. Now Microsoft is willing to make such a decision to let Xiaobing develop independently in the local market. I think it is a very good thing. I hope to be able to do better on this very good basis and give you a better surprise.