Insight into Tencent Cloud AI from a public welfare innovation challenge

Author | Xiao Man

Produced by | Leifeng.com Industry Group

How can someone who doesn’t know programming quickly develop a small program with AI capabilities?

The answer to this question was best explained in the Tencent Light Charity Innovation Challenge on May 8.

Fu Jiahui, Ding Bo, and Shi Zhenyu are contestants of the Tencent Light Charity Challenge. The team they formed, "A Half Day of Leisure in Life", is one of the teams in the "Age-friendly Accessible Design" track of this competition.

Although the three of them made it to the finals and achieved good rankings, when they first participated in the competition, they could be said to be novices in programming.

I am a civil engineer, he is an energy engineer, and he is a computer engineer. None of us has ever learned programming.

Captain Fu Jiahui said when introducing himself.

Among the entries, the "A Little Bit of Leisure in Life" team developed a small program for the elderly, which includes functions such as text recognition, rumor identification, and garbage classification.

Although the design interface and function options look very simple, the technical difficulty behind them is not low. Take text recognition as an example. The small program designed by the "Fusheng Stealing Half a Day" team needs to recognize the text in the picture and also convey it to the elderly users in the form of voice reading.

In addition, in the rumor identification module, faced with the redundant and complex information in the Internet environment, how to set up a screening mechanism through programming is also a major challenge.

However, when asked whether programming was difficult, the answer of the "Fushengtiebanrixian" team was that it was not difficult. The team later explained that Tencent Cloud's cloud development and the AI ​​technology that Tencent Cloud made available for free in this competition provided a lot of help in the process.

Tencent Cloud provides cloud functions and cloud databases for development. They provide everything you need for basic development, as well as template messages, etc. I think Tencent Cloud AI recognition is very fast and accurate, and it is very convenient to call.

It provides a master version. For example, if you want to use this product, you click on the online generated calling interface, and they will jump directly to a web page, where you can enter the credentials they want, automatically generate a master code, and directly copy and paste it into the cloud function, write the interface with the corresponding data, and then return it and call it directly to receive it. It is very convenient.

"We can take a moment to relax during busy life," team members said in an interview.

After opening the technical conversation, the "Half Day of Leisure in Life" team members further introduced WeChat cloud development and cloud functions to Leiphone.com.

The development of native mini-programs is complicated and the authentication is very complex, but the mini-programs developed by Tencent Cloud have the ability to apply for login by default, which means they can directly obtain user information. You can get the user's information by simply entering a line of code, as long as the user allows it. It is very convenient and greatly reduces our development difficulty.

In fact, it is not just the "A Little Bit of Free Time" team. There are many students and post-00s who have no programming experience, or people working in non-programming fields among the contestants. What they have are creativity and thinking, and to realize their creativity, they mostly rely on Tencent Cloud AI.

Tencent Cloud hopes to continue to promote the integration of AI technology with various industries and to continuously release the value of AI application platforms. As a commercial outlet for Tencent's AI capabilities, its strength should not be underestimated.

Leifeng.com learned that the powerful technical support behind Tencent Cloud AI actually comes from multiple top artificial intelligence laboratories including Tencent Youtu.

As of now, Tencent Youtu has more than 800 AI-related domestic and foreign patents, and more than 300 papers have been included in top international AI conferences such as CVPR, AAAI, and ICCV, which shows its technical strength.

However, Youtu's technology is not fully open on the Tencent Cloud AI platform, but needs to go through a rigorous process to output selectively and specifically according to user needs.

Generally speaking, the process of a technology moving from Youtu Lab to Tencent Cloud AI includes three stages:

• The first stage is the research stage, including market research, demand research and demand design;

• The second stage is the algorithm implementation stage. According to the research results of the first stage, corresponding data collection, data cleaning, and data labeling are carried out. When it comes to the algorithm, a good model is finally output after model training.

• The third stage is the product realization stage, which includes product design, engineering implementation quality verification and release.

From the initial research to the final launch, it actually involves many rounds of thinking and the cooperation and collaboration of many teams, and finally a truly useful AI technology product can be made.

Wang Lei, general manager of Tencent Cloud AI Application Product Center, told Leiphone.com.

Under such a long process, it takes several months for an AI technology product to be launched (excluding the research phase), which is actually incompatible with the fast iteration speed of the Internet. To solve this problem, the easiest way is to lower the quality requirements for product launch.

However, Tencent Cloud did not do so, but chose a "difficult" path - by building a large number of systems and automation tools, while ensuring high quality, it continuously improves the iteration efficiency in the R&D, testing, verification, and release processes.

"This actually took a lot of investment and energy," Wang Lei admitted, "but in the long run, when we extend the time to 2 years, 3 years or even longer, its value will naturally be reflected."

After the systems and tools were established and continuously improved, Wang Lei revealed to Leiphone.com that the average launch time of an AI product has been shortened to within one week.

At present, Tencent Cloud has launched hundreds of AI technology products. Its public cloud processes more than 3 billion images per day, more than 2.5 million hours of voice per day, and more than 100 billion sentences of natural language per day. It serves more than 1.2 billion users and more than 2 million customers worldwide.

Not only that, in Gartner’s latest 2021 "Magic Quadrant for Cloud AI Developer Services" research report, Tencent Cloud AI ranked first among domestic manufacturers in the dimension of "execution capability" (products, services, sales, etc.).

Tencent has already made plans in the field of AI technology. As of September 2020, Tencent has applied for more than 5,600 artificial intelligence patents in China, and more than 1,200 have been authorized. The total number of Chinese patent applications in the field of AI ranks first among domestic Internet companies.

Under Tencent's long-term planning, Tencent Cloud's AI technology has formed natural advantages and latecomer capabilities that are different from other cloud platforms.

The natural advantages are mainly reflected in Tencent's reserves of technology, talents and years of Internet accumulation. With the coverage of Tencent's many business scenarios, Tencent Cloud's AI technology capabilities can be polished and accumulated in a large number of database resources and scenarios before launch.

The latecomer's ability to surpass others lies in the cloud platform itself. According to Wang Lei, Tencent Cloud has a complete set of solutions for IaaS, SaaS and PaaS. Relying on the powerful computing power and algorithms of the cloud platform, the application of AI technology on the cloud platform has advantages different from other platforms.

As of September last year, Tencent Cloud AI has provided more than 200 AI technical capabilities, covering many popular AI fields such as computer vision, speech recognition, natural language processing, and machine learning.

Among them, the computer vision module alone covers face recognition, face identification, human posture, image and video processing and analysis, OCR, etc.

As AI technology is applied in the industry, the demand for scenarios increases. In addition to human face recognition, Tencent Cloud AI has also realized species recognition technologies such as pig face recognition and cow face recognition. It is worth mentioning that in this charity challenge, the mini program developed by the "Smart Fishery Team" in the "Wildlife Protection" track is dedicated to the individual recognition of white dolphins.

In addition, in order to assist enterprise users and developers in accessing application development and evaluating data and models, Tencent Cloud AI launched the full-stack artificial intelligence service platform TI Matrix and the one-stop machine learning ecosystem service platform TI-One.

In the Tencent Light Charity Challenge, cloud development capabilities were also one of the aspects that received high praise from the contestants.

First of all, cloud development is very convenient. First, it saves the backend functions. Second, it directly calls the database, saving the cost of communication between the front and back ends. Third, it is easy to call the AI ​​interface, and you don’t need to train the AI ​​capabilities yourself, you can call the ready-made ones.

Shi Moxin, a member of the participating team "Xingxing Wanwu", said in an interview.

After years of technological development and the construction of an open ecosystem, Tencent Cloud AI has begun to empower industries in many fields such as smart manufacturing, government affairs and people's livelihood, medical care, education, finance, enterprise, culture and tourism.

In addition, according to Wang Lei, Tencent Cloud AI is also actively working in the field of industrial precision engineering, which has higher technical requirements, and the field of medical image analysis, which has higher professional barriers.

To some extent, Tencent Cloud is the export of Tencent's AI capabilities and the embodiment of its technical capabilities. It is also a powerful tool for Tencent to embrace the industrial Internet.

In addition, judging from the efforts Tencent Cloud AI has made during the anti-epidemic period and its actions in public welfare events, Tencent Cloud AI is actually a mirror of Tencent's "Technology for Good" mission, reflecting the social responsibility of using technology for good.

This is exactly the best self-explanation of Tencent Cloud AI in the Tencent Light Charity Challenge.