artificial intelligence has been in the news again recently. On the one hand, Shandong, Sichuan, Shanghai, Fujian and other places have successively issued relevant policies, and on the other hand, leading companies such as iFLYTEK and SenseTime have successively released new hardware or new solutions…

There are a lot of artificial intelligence news and actions. What new progress has been made in the development of the industry, which technologies are mature, which products are sought after by the market, and which problems need continuous improvement? Let’s find out together.

“In 2018, many people thought it was the first year of the combination of artificial intelligence and industry, but the result was only the sound of thunder and no rain.” Xu Li, CEO of artificial intelligence service provider SenseTime, said frankly that the key work in 2019 is to Bring artificial intelligence to real life.

Coincidentally, Liu Qingfeng, chairman of iFlytek Co., Ltd., also believes that “2019 will be a stage in which artificial intelligence will gradually realize its dividends.” He pointed out that there are three criteria for evaluating whether the value of artificial intelligence technology can be realized: real and visible practical application cases, core products that can be applied on a large scale, and statistical application results. In layman’s terms, artificial intelligence applications must be able to “listen” when they “land”.

Computer vision matures

At present, computer vision is still the most mature category of artificial intelligence applications. In 2018, my country’s computer vision application market reached 750 million US dollars. But beyond “face recognition”, what else can computer vision do?

Take the medical industry as an example. Before, computer vision was often used for image analysis of a single disease, but now it needs diversified applications. Jia Jiaya, co-head of Tencent Youtu Lab, told the Economic Daily reporter that what they are doing now is to optimize algorithms on the one hand, and to match all types of hardware equipment in the hospital on the other, and use artificial intelligence to truly improve the medical level.

Zhang Shaoting, vice president of SenseTime Science and Technology Research Institute, said that the empowerment of artificial intelligence should not stop at recognition technology only. “For example, in bone tumor surgery, the new platform can not only identify bone tumors through computer vision, but also help achieve tumor segmentation, bone segmentation, and 3D model generation. A lot of coordination work. In this way, the preoperative preparation that used to take weeks can be compressed into dozens of minutes.” Zhang Shaoting said.

In addition, computer vision is beginning to enter the Industrial internet. For example, in the industrial quality inspection process, it used to rely on manpower, but now it can be automatically detected on the basis of image recognition. In Zhejiang Weixing Group, which is mainly engaged in the production of buttons and zippers, it used to be necessary to touch and see if the zipper was missing teeth, but now computer vision has replaced manual labor, and the 96% accuracy rate has greatly improved the detection efficiency.

In the construction industry, counting rebars on site is a routine and boring job. Yuan Zhenggang, President of Glodon Technology Co., Ltd. told reporters that this work used to require a lot of manpower to repeatedly proofread, but now using computer vision, the number of steel bars can be automatically counted by taking pictures of mobile phones. The promotion of artificial intelligence to industry management has achieved qualitative leap.

Smart assistants come to the fore

The idea that “smart speakers are a gimmick” was all the rage last year. In the first quarter of this year, the domestic smart speaker market shipped 10.6 million units, a year-on-year increase of nearly 500%, which shocked the industry.

The “counter-attack” of smart speakers stems from changes in functional positioning. Jing Kun, the head of Baidu Xiaodu Speaker, said frankly that if the smart speaker is positioned as an artificial intelligence home assistant, it can achieve more functions.

In terms of products, a major change in smart speakers is the installation of a screen. Baidu’s “Xiaodu at Home”, Alibaba’s “Tmall Genie CC”, and Xiaomi’s “Xiao AI Touch Screen Speaker” all have a touch screen, which means that in the past, smart speakers that could only input commands by voice are now many. A channel for information presentation and interaction.

“Having a screen is the first step for home smart assistants, and then cross-screen linkage and industry integration, such as the linkage of smart speakers and smart TVs, smart hardware, and even the Internet of Vehicles, etc., when the scene can be seamlessly switched between home, office and commuting. , intelligent assistants can truly realize their value.” said Li Xuechao, vice president of Tencent Intelligent Platform Products.

In addition to the personal side, on the enterprise side, intelligent assistants also move from “concept” to “effectiveness”. Intelligent customer service Netease Qiyu has just released a new function of “intelligent quality inspection”, which analyzes the content of conversations from recordings and texts based on intelligent rules, and performs secondary verification of customer service data in the form of human-machine cooperation. It seems a simple function. , but it is aimed at the “pain point” of the traditional customer service industry that relies solely on manpower and has a low quality inspection rate.

In Alibaba, the intelligent customer service “DianXiaomi” system has trained 100,000 AI trainers in traditional enterprises, and said that 400,000 AI trainers will be trained in the next three years. Lin Jingjing, head of customer service of Aokang Group, is such an artificial intelligence trainer. Lin Jingjing said that through the integration with artificial intelligence, Aokang’s customer service team has now been optimized from 61 to 48 people.

“Cognitive” applications still need to expand

From the perspective of future development, the “cognitive” application of artificial intelligence still has room for expansion, which means moving from the “perception” level such as computer vision and speech semantics to the “cognitive” level, forming new wisdom in decision-making. For example, in the smart retail industry, although computer vision can provide structured “perceptual” data, such as consumers’ age, gender, location in the store, stay time, etc., how to carry out a deeper “recognition” after obtaining this data “Knowledge” treatment still needs to be studied. From the current point of view, “perception” data is mainly used in precision marketing fields such as issuing coupons, but it provides less support for “cognitive” decisions such as product R&D pricing, shelf moving line design, and intelligent location selection. “Expanding the application scope of artificial intelligence requires a large amount of high-quality data on the one hand, and more technical support on the other hand to effectively use data to make decisions,” said Gu Hai, president of the French Artificial Intelligence Center.

How can we improve the “cognitive” ability of artificial intelligence? In terms of technology, Lu Dawei, an associate professor at Southern University of Science and Technology, believes that the combination with quantum computing should be considered. “Although it is still in its infancy, quantum computing can provide exponential acceleration effects on classical machine learning algorithms.” Quantum artificial intelligence is considered to have broad application prospects. In the field of scientific computing, more powerful energy will be released.

It is worth noting that once AI is used for “cognitive” decision-making, it means paying attention to the ethical issues of AI. According to Zeng Yi, a researcher at the Institute of automation, Chinese Academy of Sciences, the so-called “artificial intelligence ethics” refers to a series of issues such as security risks, privacy, algorithm discrimination, and supervision. Even if the machine learns autonomously, it must conform to human values, and avoid harm to other intelligent individuals through the machine’s self-modeling, self-experience acquisition and thinking speculation. Only by attaching importance to the research of artificial intelligence ethics can we lead the development of artificial intelligence in a direction that is beneficial to human beings and nature. Liu Qingfeng also said that with the diversification and complexity of artificial intelligence products, related social management, legal and other supporting construction must also be followed up.