The year before, Fu Sheng felt that the competition of the mobile Internet became more and more fierce and it was more and more difficult to move forward. Therefore, he felt that the core competition that may be the essence of the mobile Internet is over. This is his remarks on the "2017 Snowball Summit" held on March 25.
"The last opportunity seized by the cheetah is the internationalization of the Chinese Internet. What is the next opportunity? The next cognition for me is deep learning."
In his view, the essence of deep learning is to reduce technical barriers. If the cheetahs had to produce a voice product before, the face recognition was at the forefront of the world, and they could not imagine, and did not know which year to start. But because of the development of deep learning technology, cheetah has achieved some results.
The following is a partial record of Fu Sheng:
About 5 years ago, I realized that China's APP has already led the world, and China's APP can go global. The mobile Internet makes the internationalization of Chinese companies easier than before.
So the cheetah has gone all out to do internationalization. We have accumulated more than 600 million active users in 3 years, with a total download of more than 3 billion times. Each month, there are more than 600 million mobile phones running Cheetah Mobile app. 81% of the users are from overseas, and 20%-25% are from developed countries in Europe and America. We encountered some problems last year, but our growth rate is still around 25%, and the past 5 years are 100% growth.
The last opportunity seized by the cheetah is the internationalization of the Chinese Internet. What is the next opportunity? The next cognition for me is deep learning. Many people say that deep learning is a concept. Why do you want to learn deeply? In the AI+ era, Cheetah, a company that started as a tool APP, has the opportunity to make artificial intelligence and robots.
I want to see why the cheetah will have a chance from three points. The premise of these three points is that, probably in the year before, I felt that the competition of the mobile Internet is getting more and more fierce, and it is more and more difficult to move forward, so I am thinking that the core competition of the nature of the mobile Internet is over.
First, I think deep learning is an algorithmic revolution.
Previously doing voice, doing images, doing driverless, and doing spatial positioning, each is a profession. The difference in the algorithm between the majors is very huge. I found this problem by constantly interviewing the company. But one day deep learning has emerged. It can solve both voice problems and image problems. It can also solve the problem of unmanned driving. It can also solve the problem of stock trading. There has never been such a charming thing in human history. So once it is such a normalized algorithm, all the human brains are concentrated.
Deep learning, panic should be big companies, because they have accumulated a lot of technology, whether it is voice or technology, under the impact of deep learning has been subverted. Not long ago, I met a person in charge of a well-known large company laboratory in China. He said that he had done a translation for 7 years. Later, he saw a paper with keyword/keyword Google keyword/keyword, and suddenly discovered that his original technology was white. Accumulated. I believe that the essence of deep learning is to reduce technical barriers. If the cheetahs had to produce a voice product before, and face recognition is at the forefront of the world, I can't imagine, from which year do I start? But in the near future we will also disclose some of the results achieved.
Second, the algorithmic drive becomes data driven.
This artificial intelligence revolution is data-driven, and the effective growth of a paper can be very limited, but the huge amount of data generated after a large amount of data may be far better than a paper. However, you need a lot of data to try, label, and calculate, and mass-labeling data becomes a core competency. The labeling of data has only just begun. Some companies look very big, but their amount of data is very small. Today, there is a large amount of data on the Internet for people to use freely. The key path is to figure out how to label this data and figure out how to quickly generate your own data set. The game history in human history is not enough for AlphaGo to learn. The latest AlphaGo has to remove the human game because it feels that humans are not doing well. I have seen Google as a product manager for translation. He said that the new translation of this version of Google has been greatly improved. It has captured a large amount of data from the Internet and has been marked twice, but many of the data on the Internet are former. When Google translated it, they slowly removed the bad data.
Third, the opportunity for deep learning lies in the combination with the application rather than the technical output.
I believe that deep learning in the future is the basic technology application, and many companies have deep research and development capabilities. We acquired News Republic, a news product in France, to turn the user's click behavior into a labeled part of the data, and the neural network will find automatic relevant news to push. Cheetah also made a live application Live.me, now the largest third-party live broadcast platform in the United States, with hundreds of thousands of US users launching every day, generating millions and tens of millions of standard faces. This data allows us to find Accurate data. The face recognition technology developed by Cheetah has a large number of applications in pornography and child identification. This achievement is also ranked in the top of the world.
Cheetah will go all out to let AI help upgrade the Cheetah content strategy. The new mission of Cheetah is to become an artificial intelligence company with great technical ideals. Cheetah has 623 million monthly active users around the world, both capable of capturing big data and capable of tagging data.
China is in the forefront of this wave of technology. 40% of the papers in deep learning were published by Chinese. This time we communicated with experts without language barriers and no time difference. And today, the computing power of Chinese GPUs has accumulated over the years. I believe that the mobile phone era will surely come. The use of mobile phones on mobile phones should not have explosive growth, but more and more smart devices like Amazon's echo and our dream robots will definitely enter our lives. . We have good data, a huge sample group, a good team of engineers, and world-class manufacturing capabilities. The biggest opportunity in China comes from robots, and the era of coexistence of people and robots will surely come.
Dongguan Yijia Optoelectronics Co., Ltd. , https://www.everbestlcdlcms.com