Shulou(Shulou.com)06/03 Report--

From October 17 to 19, 2018, the 10th China system Architects Conference (SACC2018), hosted by IT168's ITPUB Enterprise Community platform, will be held in Beijing. This conference includes four main lines: core business system architecture design, big data platform architecture, traditional enterprise digital transformation, and artificial intelligence application.

In order to give you a deeper understanding of the schedule setting and special benefits of the SACC2018 conference, the ITPUB community & IT168 enterprise editorial department conducted a series of interviews with more than a dozen producers and expert consultants.

Guest interviewed in this issue: Li Dahai, partner and senior vice president of Zhihu, main line 4: artificial intelligence application landing practice (part two) special producer.

Li Dahai, who has many years of experience in search algorithms, joined Zhihu in 2015 and has been in charge of advertising technical team, data, algorithm and overall community business. Currently responsible for Zhihu big data team, content circulation and AI new business development, in the related technology is a rare talent.

Zhihu, a company with very Silicon Valley style and geek style.

Zhihu, which was officially launched in 2011, is a national knowledge content platform, which maximizes the value of knowledge, is used by everyone, and achieves universal knowledge. Li Dahai, senior vice president of Zhihu, said that Zhihu is a very Silicon Valley and geek company with flat management and a pirate culture that is "always manageable." Zhihu condenses the development and values of the company with the self-growth of employees to do interesting and challenging things together.

The increase of 60 million users in five months is not covered by Zhihu's latest AI algorithm application practice.

It is understood that by the end of March this year, Zhihu had 140 million registered users. After five months, the number of registered users of Zhihu has reached 200 million by the end of August, and other figures are also growing rapidly.

Li Dahai said that at present, Zhihu has applied the AI algorithm throughout a number of scenarios from content production and consumption to user connectivity and community governance. In the content production scene, Zhihu automatically generates problems through the application of the AI algorithm, enriches the content dimensions, and quickly distributes the content to the appropriate users to stimulate their desire for creation. In the content consumption scene, Zhihu uses the home page recommendation system "Crystal Ball" and search system to help each user get more interesting and useful content quickly and easily. In the user connection scene, Zhihu comes online with a series of functions, such as user follow recommendation, similar responder, user clustering and so on, to strengthen the connection between users and promote the creation and sharing of content. In the content governance scene, Zhihu constantly optimizes the algorithm robot "Wall-E" by comprehensively using emotion model, user intimacy model and text recognition model to identify and deal with unfriendly, unanswered questions, strange mood and other contents in real time. The accuracy of some scenes reached 99.13%, creating a professional, serious and friendly community atmosphere.

AI is not just a gimmick, the real combination of problems to meet the demand is the real landing.

The landing of AI applications is a clich é, and people have seen gratifying developments in the past year or two. Li Dahai said that there are many factors that determine the landing of AI applications, but there are several key points: first, whether there is a large amount of high-quality data; second, whether there is an effective AI talent training mechanism; and third, whether there is the ability to integrate closely with the business.

To judge the landing of AI applications, we should also look at these three points. AI is not just a gimmick, really solve the problems around people, meet everyone's needs, is really called landing.

Is the algorithm biased? Man-machine combination is one of the effective means.

The popular "algorithm bias" in the industry says that AI algorithms and their decision-making procedures are shaped by developers, and any imbalance in the written code and subsequent training using training data will have a great impact on the results ("eight phenomena demonstrate that artificial intelligence threat theory really exists"), so people's affirmation of AI is erratic.

Li Dahai said that Zhihu promotes AI applications in order to build a platform that conforms to Zhihu community and content values. However, it does not mean that AI will completely replace human labor. Zhihu's AI application is accompanied by a high degree of human participation from beginning to end. Man-machine combination is one of the effective ways to avoid "algorithm bias".

When formulating community rules and specific applications, Zhihu's technical team will continue to train algorithms to correct possible deviations in order to pursue the positive effect of AI applications.

AI algorithm practice two big pits, do not continue to "alchemy"!

The practice of AI algorithm is a complex process, in which there are many "pits" that need to be paid attention to. Li Dahai cited two classic examples, hoping to be helpful to practitioners.

First, there is too much division between "algorithm" and "architecture". The students who do the algorithm do not care about the implementation at all, the efficiency of doing so is relatively low, and it is not conducive to the landing of the algorithm.

Second, it is too shallow. Some so-called "algorithmic engineers" do all kinds of models and structures blindly, randomly adjust the super-parameters according to the effect, and whether the effect is good or not depends entirely on providence. This kind of engineer is more like the ancient "alchemist." maybe the final output can be used, but the engineers themselves will not grow very well.

Artificial intelligence has entered the landing stage of wide application, please check a "blessing bag" of practical information!

The concept of artificial intelligence has been discussed for many years, and finally entered the stage of widespread application, which is also the core of the whole industry. The China system Architects Conference (SACC2018) provides an opportunity for colleagues in the industry to share their experiences.

Li Dahai, as a special producer of SACC artificial intelligence application landing practice (part two), said that in this special show, we have invited several companies with strong technical strength in the industry to share with you examples of their landing in their own business. It is hoped that their speech will bring some new perspectives and broaden ideas for the participants, and that this conference will serve as a link to connect colleagues in the industry, communicate with each other on an equal footing and exchange needed goods.

The special schedule of SACC2018-- artificial intelligence application landing practice (part two) is as follows:

Ticket entry: http://sacc.it168.com/goupiao.html

The official website of the conference: http://sacc.it168.com

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.