AIR 022 Tencent millet 360 today s headlines in the eyes of the machine learning

Unstoppable tide of machine learning, technology giants how to sail? CCF-GAIR the General Assembly first discussion forum on the topic of a heated debate. Microsoft Research Asia Director featuring host researcher Zheng Yu, millet, co-founder of Mr Huang Jiangji, 360, artificial intelligence Research Institute President Yan Shuicheng Dr HUANG leaps, Tencent founder Dr, today’s headlines scientist Dr LI to participate in discussions. (Deletion)

Advance of Giants

Zheng Yu: share 360 what progress had recently been achieved in the field of artificial intelligence, and what kind of layout in the future.

Yan Shuicheng: 360 before I joined in artificial intelligence has already done a lot of things, Basic Visual, sound, meaning there are large data. In addition to the traditional business, has had two operations on the use of artificial intelligence, an intelligent hardware, a favorite old week live.

From a research perspective, we do two things: one is when we have a big data more quickly trained the model? From intelligent hardware level, in addition to the traditional artificial intelligence, used in the cloud there is a very important scene is the end of the artificial intelligence.

Side there are two levels of artificial intelligence can do things, one was you better model in terms of how algorithms, lower complexity, with very few resources on the use of artificial intelligence to end well; the other is from the surface, such as promote deep learning chip’s computing power. But China’s specific situations, users on the value of the product are, in my personal view, from algorithm reduces the complexity of the calculation are very valuable, so over a period of time we have to spend time and effort very much.

Layout for the future, I think there are two aspects: one is the operational level, security operations continued to hover around 360 core gene, using artificial intelligence to provide a service. From a research perspective, we will continue to focus on building better models, algorithms, enhance the ability of artificial intelligence on the side and on the ability to use data to increase the depth of learning.

360 artificial intelligence Research Institute President Yan Shuicheng

Zheng Yu: the second issue I would like to ask Dr HUANG leaps, Tencent, the audience on the sector aren’t very well understood, would you please introduce this sector, as well as what you do in the field of artificial intelligence and its characteristics.

Yellow leap: figure is now owned by Tencent Department, the team was founded in 2012, initially is a social networking group following a team, and today we have spontaneous growth as a Department, we have done and algorithms of artificial intelligence research and development and business ground, now has dozens of advanced technology and the ability to scale of hundreds of millions of images … …

Graph is defined to do related research and development, but all of our research and development is based on the actual needs of the business, which will then form a relatively good business, ground loop, you can instantly know the good and bad of our business, update our algorithms.

Teams now more optimistic about the direction there is recognition and is mainly engaged in business related fields. Such as biometrics, we are relatively optimistic about the first face recognition and voice recognition, then the fingerprint and other identification: recognition, image and voice recognition, and so on. Level I think face recognition technology is relatively mature, believe that in the next few years face authentication have the opportunity to move from Internet to offline, such recognition is now face recognition have greater value. Such as image recognition, its concepts and more extensive areas, I believe image recognition in later years would have an explosive growth.

The latest augmented reality, chat bot technology is I am more positive, these technologies improve human life experience, and it can and has a very good combination on the line below the line, closer, more likely to have a more realistic value and results.

We hope that by using artificial intelligence technology really change the quality of life for Internet users. Meanwhile, we will also put some of the latest research and development results through the development of platform and Tencent cloud output, ensuring technology can really improve the quality of life for Internet users.

Tencent founder Dr HUANG leaps

Zheng Yu: Li Lei is a graduate of Chiao Tung University began to go to Baidu, now, in today’s headlines, from your personal point of view in the artificial intelligence field, Baidu and what are the advantages and disadvantages of the today’s headlines, especially insufficient to clear.

Li: before I got the headlines in Baidu’s deep learning Academy for two years, work basically revolves around deep learning and spontaneous speech understanding, to I headline after headline laboratory, work is built around deep learning, natural speech understanding and learning, before, I now do much more on the content.

I think most of the two companies have in common is that they are artificial intelligence companies. Baidu no doubt invested very large in artificial intelligence, voice recognition and drone are representative of all conventional wisdom that today’s headlines is a news client, but now the headlines is a platform for sharing all information, content creation. One side of the platform are content creators, content consumers on the other side, in order to recommend the best content for readers in need, need is a machine learning technique. From this point on, the two companies are based on artificial intelligence technology as the core of the company.

Difference lies mainly in the development stage and the volume. Baidu and Google, Microsoft, unison, has already begun studies in this field: Microsoft in 2009 started speech recognition; Google in 2012 with 1000 computers, 16,000 CPU to do a cat face recognition; Baidu to do deep learning Institute, do voice recognition and image recognition, you can do almost all areas of artificial intelligence.

Today’s headlines and the companies are not the same, since 2012 has only 4 years, we can think of 4 years when big companies do? Facebook Research Institute; Baidu no laboratory; Google search. Today’s headlines in 2014, when the whole team are said to be at the forefront of the consciousness did a great deal of spending on technology. On this frontier technology in artificial intelligence, whether now used, or is doing cutting-edge research in order to technical reserves, today’s headlines on this point is very, very good idea.

Today’s headlines scientist Li Lei

Zheng Yu: many of millet are artificial intelligence, but we want to know what is millet is not artificial intelligence, is very bad? Because we have a lot of small business representatives, where you do well could be their chance. You can say what you want to do, but can’t do it well, it’s hard to do less than what?

Huang Jiangji: that’s a very good question. A lot of people say now is the spring of artificial intelligence, for many startups are overtaking opportunity, and I think this is a fact. But in fact the corners and any corner in the past have not been substantially changed business more difficult. I think if you want to venture a success, you still need to what is a deep understanding of user pain points and how to get your products to better meet customers ‘ pain points. Not just mastered a new technology, you can naturally acquire mass users. I think today’s headlines is a good case, continue to use new technologies to solve users ‘ pain points. This is my understanding of the startup opportunities.

Answering the question raised Zheng Yu, I do not think that millet does relatively bad in the field of artificial intelligence which, I think we should ask is where we present our products features, user experience and overall product satisfaction, which are bad, and I think that these are opportunities. If you close the door to ask me, I think all of our products are doing enough intelligence, we have to do is tip of the iceberg, there is still a long way to go.

Millet is co-founder of Huang Jiangji

Academia and industry

Microsoft Research Asia competent researcher Zheng Yu Moschino Galaxy S5

Zheng Yu: many colleges and universities teacher or in research institutions who want to go to the industrial sector, but there is a lot of apprehension, I went to after can survive, so I want to ask a provocative question, do you think 360 artificial intelligence which restricted or where you are not satisfied with?

Yan Shuicheng: I General explained problem like from martial arts of angle starting, I think in academia, artificial intelligence of research of method is like, hope you is a family decent, beat has others, also can refined some moves out, these moves in and others fighting Shi can won others; in industry this moves became not so important, important of is you has a martial arts, with a cooperation of way put others beat, put a problem solution off, this is good of artificial intelligence.

Issue is you have to have artificial intelligence ideas and methods to solve this problem. I join the 360 is because he has his advantages, in particular times, our network can be said to be leading academics in the field of artificial intelligence, artificial intelligence resources to best advantage. Moschino S5 Case

When I add 360, smart hardware is a critical business, now live is very critical to the business, which is exactly what I wanted to do. If there is any malpractice, is the pace of the Internet much faster than academia, in academia, for example six months or three months of good results came in, to have a good academic paper, but this is obviously not suitable for industry.

Zheng Yu: more is to look at the future of the Institute, bosses aren’t specific targets?

Huang Jiangji: Microsoft Research Asia is not standard, is certainly a good Research Institute in China for two-legged go hand in hand. What results when you have the need to support existing businesses, algorithms of artificial intelligence or play a role in the business. In addition you need to have a long-term idea, 1.5 years or so a year later what kind of artificial intelligence technology or to hatch their own business, but in a very short period on the line. This is a good way to ensure Research Institute within the companies survive.

Progress and development

Zheng Yu: AI in the voice and vision have made a lot of progress, but not much progress on the text, technical direction in how to solve this problem?

Huang Jiangji: I of feel is with depth learning in processing natural language and natural language understanding aspects made of results is far than yiqian more has, and this year almost 2/3 above of papers is with depth learning to do of, so just that a premise said we seems no see depth learning in natural language aspects like voice technology and image technology as has is big of success, actually not, just late has some, now has is big of progress.

Yan Shuicheng: personally I do image and data are much greater, for natural language understanding is not particularly large, how people process language the piece, with the way we now use but it is still not the same. From the perspective of people, now the whole direction, there are essential differences in the way with people, there is a great deal of change. Other hand, from the perspective of image, if you want to do the general nature of this image search but it is still very difficult, whether it be Google or Baidu.

Zheng Yu: Finally, each expert in one sentence summarize the development of artificial intelligence.

Huang Jiangji: my opinion could change a machine learning engineer product development methodology, upside down, just completely different.

Yan Shuicheng: artificial intelligence in academia and industry who have a common language, this is very good.

Yellow leap: Please do not equate AI machine learning, machine learning does not equal the depth of learning. We should not just in our industry on an issue, and we want to solve the problem more complex than a supervised learning, we need to get the computer to learn a few simple questions, but also to let the computer learn to logical reasoning, learning how to interact with environment in a complex environment, while also learning how to use the least amount of energy to learn new skills.

Li Lei: a Word is to moisten things silently, AI is like more than 10 years ago the Internet into our life, it is able to integrate into our daily lives, can improve our quality of life, and artificial intelligence are not aware of our people everywhere.

Like this:

Like Loading…

Related