(Original title: The first year of domestic AI commercialization: three major application scenarios need to be landed)
On June 21, the “Forbes†article pointed out that TOPBOTS, a well-known tech media in the field of artificial intelligence, selected 20 technology leaders who are driving the reform of artificial intelligence in China, Li Kaifu, co-founder of Innovation Factory, President of Baidu Group and COO Luqi, and Tencent. Many senior professionals from the field of artificial intelligence in the industry and academia, such as the director of artificial intelligence laboratory Zhang Jian, have become the "Chinese competitors" most concerned by the West.
One of the most essential reasons why AI talents are so concerned is that 2017 has become the first year of global commercialization and productization of artificial intelligence. Unlike the era of machine-building, the AI ​​commercialized the era and the two countries can start simultaneously.
According to the MIT Technology Review, the world's top ten breakthrough technology list released in 2016 and 2017, we can see that in addition to Google, Microsoft, Apple, and other traditional American strong technology companies, the Chinese AI market has also risen. Taking Baidu as the starting point, Alibaba, HKUST News, Face ++, etc., are the initial pattern of “super-strong and strongâ€.
In fact, domestic and international AI commercialization “scratch battle†has only just begun. The pattern is far from a foregone conclusion. “Technology promotion + application landing†is the key to the AI's move from the laboratory into the homes of ordinary people. China How does the company fall?
One sign that has been seen is the human-computer interaction with "smart speech operating system" as the core, intelligent image recognition with "face recognition" as the core, and the machine brain with "smart driving" as its core, and the behind-the-scenes Connected home scenes, transaction and identification scenes, and car scenes will gradually become the main battlefield for the commercialization of AI.
Human-machine dialogue "Tear" open scene: Voice operating system is coming
Speech recognition has evolved into a pioneer in which artificial intelligence is exposed to the public. Existing products that are closest to ordinary users appear to be voice input methods, Apple Siri, and Amazon Smart Sound Echo.
Just two weeks ago, HKUST had just released the MORFEI microphone and the AIUI open platform, and the stock price rose rapidly. As early as 2015, HKUST had already applied its microphone array and voice technology to smart hardware. It cooperated with JD.com and launched the first smart speaker in China.
It is worth noting that, in addition to smart speakers and other hardware, 2C direct commercialization is not easy, many manufacturers began to explore the ability to package output, first 2B and then 2C.
For example, following last year, the Tencent Cloud United WeChat AI team launched an intelligent voice service, and recently released a smart voice solution for voice recognition, speech synthesis, and voiceprint identification, providing vertical industry-customized voice products for the industry. Little micro."
Baidu, who proposed "AI first," has a voice recognition rate of 97% and was selected as one of MIT's "Top 10 Global Breakthrough Technologies." Since the beginning of the year, the wholly-owned acquisition of human-computer interaction startup team Raven Technology, and upgrade the secret team to the Department of secret affairs.
Li Yanhong, the founder and CEO of Baidu, said on many occasions that artificial intelligence products combining hardware and software will be the new direction. Baidu is accelerating the process of exporting artificial intelligence technologies to smart homes and other fields.
When acquiring Raven Technologies, Lu Qi stated in his internal e-mail that this is an important step for Baidu to lead the future development of artificial intelligence. It is of great significance for establishing the leading edge of Baidu's intelligent interactive platform and building the core competitiveness of hardware and software integration.
In the first half of this year, Baidu and Xiaoyu jointly released the latter's new product “Split Fish†video calling robot, which is Baidu's dialogue-based artificial intelligence operating system, DuerOS.
It is understood that DuerOS is affiliated with the Baidu Department of Secret Business. Apart from the small fish at home, the system has already reached cooperation with well-known companies in various fields such as CITIC Guoan Guangshi, Midea, Haier and Lenovo. These cooperations precisely reflect the logic of Baidu's outbound output of DueerOS - widespread network expansion.
Another example is that at the end of March, Baidu released its DuerOS smart chip and announced that it had reached a strategic cooperation with Ziguang Zengrui, ARM and Shanghai Hanfeng. According to reports, this cooperation will build a three-tier structure that includes secret brains, voice solutions, and chips/modules. Among them, the first two layers are provided by Baidu Ding Mi, which gives the chip DuerOS the "core of dialogue" core functions. The chip module blocks are jointly supported by Ziguang Zengrui, ARM and Shanghai Hanfeng.
The industry believes that the release of DuerOS smart chip will expand intelligent voice interaction capabilities to more intelligent hardware, which can further reduce the development threshold of smart hardware. In addition, the introduction of artificial intelligence to traditional manufacturing industries is expected to help upgrade the industry and improve the ecology of the smart Internet of Things.
When image recognition is in progress: China has been able to “sweep the planeâ€
A technology from development to application will have a time difference and breakthrough. For image recognition, face recognition is currently one of the main directions for users.
At present, there are many startup companies in the market that use face recognition as the direction. One of the representatives is the Face ++ (Technology), which was selected as the “Top Ten Breakthrough Technology List of MIT Technology Reviewâ€. The list stated that Face ++ pioneered a new usage scenario of face recognition technology, from cracking fraud cases to “smile paymentâ€.
In addition to some startups, in the domestic market, another large company that has long studied face recognition is Baidu. And, in the two authoritative international evaluations of this technology, FDDB and LFW, Baidu has won the first place, among which the accuracy rate is 99.77% in the evaluation of LFW.
It is not easy to have this good result. The difficulty in researching face recognition technology is different from ordinary image recognition. First of all, in terms of human facial features, each person's facial structure is similar, which is unfavorable for using human faces to distinguish human individuals. There are special circumstances, such as twins or even multiple births.
Second, it is the influence of external factors such as expression, lighting conditions, and cosmetic surgery. Different facial expressions, angle observations, influences of lighting conditions, and face masks such as masks, sunglasses, hair, beards, and even cosmetic surgery and P-characters all increase the difficulty of face recognition.
However, face recognition, as the core technology of Baidu image recognition, can already achieve face detection, face comparison, face search, and is used internally in products such as Baidu Magic Map, Baidu SkyDrive, and Baidu Insight. For external applications, Baidu has entered into cooperation with Taikang Life Insurance to apply face recognition technology to the return receipts and return visits of the latter’s WeChat insured, and to verify customer identities online.
One detail is that at the Baidu World Congress venue last year, there were no obvious facial recognition areas. A few people found this area, and after they took pictures at the scene, they quickly entered the field from the face recognition channel. Most of the participants attended the conference. The players are still lined up for a long time. However, at the Baidu Union Summit in May this year, everyone had chosen “brush face†admission.
From a small test to a big show, Baidu began to push face recognition technology to different industries.
On June 28th, China’s first personal face recognition intelligent boarding system, Nanyang Jiangying Airport, landed on the ground. Although it was only a regional movement, it meant that “sweeping face boarding†had become a reality.
According to Lin Yuanqing, president of the Baidu Research Institute, the airport is different from scenic spots and company access control, and it has extremely high requirements for safety. Baidu face recognition technology lands on airports, and it is hoped that advanced artificial intelligence technology will not only ensure the security of security clearance, but also To further enhance passengers' traffic efficiency and reduce the pressure and heavy workload on the airport's human resources.
In fact, Baidu’s face recognition technology has been launched in Wuzhen, Wuyishan and other tourist attractions. Through the accurate face recognition of the gates, the experience and management efficiency of the park have been improved. It has also cooperated with the nationwide parenting platform “Baby Homeâ€. The face recognition system compares and evaluates and assists the family to quickly find the information of the lost person.
In the face recognition competition of the “Most Powerful Brain†program in the fourth quarter of Jiangsu Satellite TV at the beginning of the year, on behalf of Baidu, a small-scale robot won the “biggest brain†of mankind.
The future of automatic driving: "open empowerment" and more people entering the game
“Electricity, intelligence, and networking are the future of the car. We recognize and will lead such technology trends.†— A word from the Weilai car official website points to another important area of ​​AI's future smart driving.
In the middle of this year, Weilai Motors, one of the representatives of the research and development enterprise of the first-in-class smart electric vehicle, announced the C-round financing. The list of investors, in addition to a number of investment institutions, also includes Tencent and Baidu. This actually reflects the two camps of the current auto-pilot, one is the intervention of the Internet company is called the break, the other is the traditional car companies must bet on the frontier direction.
At present, technology giants including Google, Tesla, and Apple all continue to invest in autonomous driving. Ford, BMW, Mercedes-Benz and other large car manufacturers have also applied primary autopilot technology to their own cars.
Baidu is a special one in the list of company companies that have a list of layouts. Because it invests in foreign investment, it is also actively exploring autonomous driving.
Gained a broad understanding of Baidu’s actions in this business. It has started large-scale research and development of unmanned vehicle technology since 2015. In December of that year, it conducted full-automatic driving tests of highways and urban roads in Beijing; it was acquired in September 2016. Self-driving road test license in California, United States, in November in Wuzhen, Zhejiang Province, the ordinary open road unmanned vehicle trial operation.
In April this year, Baidu announced a new plan called Apollo, which announced that it will provide partners in the automotive industry and autonomous driving a complete set of software, hardware and service solutions, including vehicle platforms, hardware platforms, and software. Platform, cloud data services and other four parts.
When can automatic driving be brought to the market? A Boston Consulting Group study believes that the market for self-driving cars in 2025 will reach $42 billion. By 2035, a quarter of the world's cars will be self-driving cars.
If you are concerned about the scenarios in which automated driving may be used first, cooperation with the travel platform has always been one of the directions discussed in the industry. The reason is that compared to letting ordinary people make up their minds, trying to buy a driverless car and catching up with platforms such as Uber and intelligent traffic is a relatively feasible initial route.
Because of this, in May 2015, Uber poached 40 researchers from the Carnegie Mellon University Robotics Institute and set up an unmanned research department; in July 2016, it acquired a company that was only established for 7 months. Self-driving truck startup Otto.
As a travel platform, Uber's involvement is inevitable, but beyond the potential advantages, Uber is now faced with the problem of team restructuring, but also to deal with Google's unmanned company Waymo to its lawsuits for stealing driverless technology.
Just this month, The Information scrutinized the global research and development of unmanned companies and ranked them by three major factors: technology, engineering (talent) and business model. In conclusion, Waymo ranked first, followed by Uber, which has been turbulent recently. Baidu ranked seventh and ranked in front of Apple. Didi also entered the list for the first time with its new automated driving program.
Coming of "AI+" and "Soft and Hard" Era
In fact, if we look at the entire global AI market, Google, Amazon, and Apple have gradually introduced or updated smart speaker new products since this year, and the commercialization of the voice field has started to land.
Caixin Weekly pointed out that in the field of intelligent voice human-computer interaction, the three largest AI technology giants in the United States have already presented two modes: Google and Apple assume that users have mobile phones, TV set-top boxes and other products besides smart speakers. They only need to By connecting each device to the software, a screen can be added to the voice assistant at zero cost. Amazon, on the other hand, integrates software and hardware. Whether or not users have bought related products in the past, they can directly shop on the Echo Show.
Returning to the domestic market, companies such as Baidu, HKUST, and other companies that have been laying out in the AI ​​space have seen their AI capabilities through the cloud and chips used in televisions, refrigerators, set-top boxes, speakers, and wearable devices. Although the commercialization of AI in the “speaker†category is less than that of the United States, in other areas of “soft-hard†integration, and in areas such as using intelligent voice systems to search for information, takeaways, movie tickets, or even payment, O2O is better than O2O. There is an advantage in the inadequate US market for mobile payment layout.
Another clear line is that domestic Internet giants such as Baidu are trying to export AI capabilities through open platforms, accelerating Chinese commercial applications of artificial intelligence technologies in different scenarios and industries, which will give the entire Chinese traditional industry Smart transformation really invigorates.
One notable detail is that on July 5th, the first Baidu AI Developer Conference will be held soon. Baidu will release its overall strategy, technology and solutions for its artificial intelligence open platform for the first time. Perhaps, as Baidu COO Lu Qi said: "In the era of artificial intelligence, the success of developers is the real success of Baidu." The entire Chinese market after being empowered by AI may also emerge from the new road of technology and manufacturing overtaking.
(Text / Zhao Na)
Dongguan Guancheng Precision Plastic Manufacturing Co., Ltd. , https://www.dpowergo.com