Later, some people began to realize slowly that the singularity theory was indeed exaggerated. But does AI have any impact on our lives? There are indeed many, if not many, times when you can feel that some skeptics are not exaggerating. Some people even recently said that as people rely more and more on machines based on big data and various algorithms to help them make It was decided that we have actually given our robots control over our lives.
In this issue, we invited Dr. Hong Xiaowen, Dean of Microsoft Asia Research Institute, to explain to us whether AI's rapid development can be compared with our intelligence and talk about Microsoft's development strategy in artificial intelligence (below). .
Xiaowen Hong, Senior Vice President of Microsoft, Chairman of Microsoft Asia-Pacific Research Group, President of Microsoft Research Asia. We have made outstanding contributions to the development of Microsoft's SAPI (SpeechAPI) and SpeechEngine technology, and have won many awards and awards in different categories. It is also an IEEE Fellow (IEEE Fellow) and an internationally recognized expert in speech recognition. He is currently the editorial board member of the Communication of the ACM and has published more than a hundred articles in internationally renowned academic journals and conferences. Academic papers. His book Spoken Language Processing, which he co-authored, has been adopted as a phonetic textbook by many universities around the world. In addition, he has 36 patented inventions in various technical fields. Hong Xiaowen has not only been responsible for basic research but also managed product development. At the same time, he has also been involved in various research and development fields. Dr. Hong Xiaowen is the chairman of Microsoft Asia Pacific R&D Group and president of Microsoft Asia Research Institute. He is currently responsible for promoting Microsoft's research and product development strategy in the Asia-Pacific region, as well as cooperation with China and the academic community in the Asia Pacific region. He believes that his "dual perspective" formed over the years may help young domestic scholars to grasp the correct direction.
From 2006 to the present, there has been no major theoretical progress in deep learning. However, the enthusiasm of the industrial sector is now rising. What are we promoting blindly?
Everything has two sides. Deep learning or AI is indeed very hot now. As the moderator said, there is currently no major breakthrough in theory. The reason for this is because when a theory has a subversive progress, its application threshold is reduced.
Today's deep learning has been given a mysterious color, which is the theoretical development of its creative development will reduce its application threshold, and now anyone as long as they have the opportunity to use open source programs and big data, can achieve a good application. From this perspective, this is why deep learning is welcomed by people in the industry.
On the other hand, deep learning certainly cannot solve all problems. When you have big data resources, deep learning can make the best decisions, but it does not know why these decisions were made. At present, there is no problem in deep learning and dealing with machines. However, when dealing with people, if you do not get people's convincing, in many cases, deep learning will not be so useful.
Another is creativity. Today, so-called deep learning, machine learning, and big data are 95% cross-cutting. One of the greatest is creativity, and creativity is a small data problem.
For example, a topic that is very popular this year is called gravitational waves. Think about today, 100 years later, we need to use very expensive equipment to be able to measure the existence of gravitational waves. Then, how did Einstein propose gravitational wave theory 100 years ago without data? I think This is a problem that deep learning cannot answer.
Here we can make a summary:
From 2006 to now, when a disruptive innovation emerges, it will have a major impact on human life, but this means that the next innovation may take more time. For example, anyone who studies mathematics or physics knows that if you solve a problem today, it does not mean that the next problem will take a shorter time. This is not like Moore's efficiency. This is not understandable from the perspective of human development. So, I think this is a very natural phenomenon. When you have a disruptive breakthrough, the next innovation will take more time and brainpower. Historical experience tells us that this is an irrefutable truth.