Artificial intelligence (AI), or deep learning (DL), has been talked widely and attracted billions of investments during the past years. From a technology in the lab to real-life application, AI’s landing ability has been a question for numbers of AI startups’ founders and investors who believe in the future powered by AI but are troubled by its application. The development of AI in security-related area demonstrates a promising future of AI, at least in this industry. AI believers are fueled and have more confidence in what they believe.
Security becomes the best playground for AI commercialization. The driven power to make security the best landing scenario of AI technology could be deducted into three aspects: the society, the technology, and the industry.
China is a country with more than 1.4 billion population from 56 ethnic groups. The number of populations doubles the difficulty in administrative management. Along with the urbanization process in China, the situation becomes even more complicated. According to the National Bureau of Statistics of China, the urban population in 1982 was only 20.4% of the total population. The urban population reached 57.4% in 2016, and urban citizens were more than 792.9 million.
The large growth of population is a critical indicator of large city disease and increase the security management difficulty and cost of the city. The development of the city urges security to upgrade. In the 1980s, China started to install security cameras. The video taken by security cameras started to contribute to current security database. The accumulation of security data increased geometrically due to the ascending number of security cameras and other security data input devices. Increased input resources and population growth, security management demands a better and more efficient solution.
Besides the population, the policy is another trigger to push the intelligentization of security to the next level. In 2015, the State Council of the PRC published the guide to enhance the construction of security system and complete the security management and danger prevention mechanism. Several policies published since 2016 advocated intelligentization process of the security system and encouraged security enterprises to cooperate with AI technology companies to develop smart security product and solutions.
Smart City plan is common in a lot of countries’ strategic plan, so is China. Smart City includes digitalized government service, digitalized industry and business, and digitalized citizen life service. Security is an important component of Smart City and is thus accelerated by the construction of Smart City.
With the depth and width of AI application and the development of integrated circuit, computation capability, and algorithms, it can be expected that the pan-security industry will be equipped with AI technology within the next five years.
Technology can be considered as the foundation of AI’s development and so for smart security. In today’s world, technologies are connected, and they influence, promote, or even hinder others’ developments. There is no technology can isolate itself from other technologies’ support. AI is an integrated technology, which is heavily attributed to the development of computer science and data science.
The R&D process of AI algorithms require massive data to do the training and repeatedly doing the training and improvements/innovations on algorithms. Computer vision (CV) is a typical vertical of AI technology that demands crowds of data. AI application in security relies on CV heavily. For instance, the image captured by the camera will be analyzed through facial recognition technology to identify if any abnormal activities or suspects at large for the police department.
Besides facial recognition, other CV verticals also contribute to this-and-such scenarios. The update of equipment is an ongoing process and the quality of the image provided by different input device varies depending on the usage year and model of input devices. The difference in image quality affects the turnout generated by CV and therefore image processing (a vertical of CV) aims at optimizing the image quality to turn the poor-quality image into a recognizable one. However, image processing can also process a high-quality image and reduce or even eliminate image noise and unnecessary or unrelated information.
Other than CV’s development, cloud computation and telecommunication also help the intelligentization process. Before 4G or even 3G era, cloud computation’s ability was limited.
Traced back to 1980s, the density of cameras in a neighborhood did not require much manpower to analyze the tape. However, the definition of the image captured was the main focus in upgrading – people wanted high definition (HD) video and hence formed the first wave of the upgrade, the HD upgrade.
Entering the 21st century, internet reshaped people’s life and also industries. Security was thereby welcome the second wave of the upgrade, the internet era. The change brought to the security industry was that data recorded for security purpose was transmitted back to and stored in the data center. The benefit from the centralization management provides convenience and broadening security coverage.
The AI technology is believed the third upgrade for the security industry. With internet development and the enhanced sense of security, the security data began to boom. The massive data and low efficiency in data processing done by the traditional way (manpower) could not sustain a healthy growth of security management.
The intelligentization of the security will release man from redundant works and promote process efficiency. CV is the most used AI technology in the security area. The maturity of CV is able to serve for the security purpose. Face recognition, a branch of CV technology, has been widely used in different scenarios for identification services.
Applications in Security
There are several AI technology companies doing business in the security area: SenseTime, Yitu, Terminus, and so on. SenseTime is considered the most successful one in AI security and it claimed in 2017 that it had reach breakeven point completely. Cooperating with governments, SenseTime provides integrated smart city solution. By now, SenseTime is valued at USD 6 billion after receiving its latest financing in 2018.
If divided by landing scenario, security can be dissected into housing security, transportation security, public security (i.e. campus, park, etc.), and industrial zone security. The logic to plan landing AI into security should start from ToB (To-Business) and even ToG (To-Government), who have more demand and the urge to obtain smart security solution to improve efficiency and accuracy and/or lower the costs in a long run.
Individuals do not have the incentive to obtain smart security solution for personal or family use since China is the safest country in the world. ToC (To-Customer) business will be a hard bone to bite for an AI technology company at the very beginning. ToC business’ success relies on satisfying the unsatisfied need and acceptance. AI is not an unfamiliar concept for investors and technology area, but it is a new thing for the mass. The cost of culturing the mass about AI, elevate the acceptance of AI, and persuade individuals to use AI products are too high for a newborn industry and a company.
Rokid and ToC Market
Rokid is an AI startup who targets at ToC market with featured smart speakers. The smart speaker is the connector between the individual and the smart home. In the overseas market, the popularity of smart speakers like Echo and Homepod is because of the market acceptance and the culture with strong technology property. China’s development is at a fast pace and the mindset of citizens is changing drastically, but the mindset change is among generations not in the same generation.
The young generation is more willing to buy smart devices like a smart speaker. Rokid’s smart speakers are attractive to the group instead of the 40s and above. However, the smart speaker is a competitive market. Rokid’s smart speaker might be smarter than other smart speakers in the market, but a smarter speaker’s prevalence depends on a higher smart home level of the society. Unfortunately, China is not a mature market for smart speakers to demonstrate its strength. Taking smart lock as an example, South Korea’s smart lock market penetration rate is more than 90% and the west and the U.S.’ data is approximately 50%, but China’s smart lock penetration rate is below 10%.
The smart speaker’s “smart” function is weak in a smart home setting in China, and it is more like a “Siri”, a voice assistant. In this case, Rokid’s AI strength is not acknowledged (or has no chance to be testified) by the market since there is no proper or enough landing scenario (smart home). There are other smart speakers in the market like Tmall Genie, Xiaomi AI (小爱), Myna AI, etc. Smart speakers have similar performance in being a voice assistant and the price and acoustic quality are main indicators for customers.
Speaking of price, Rokid does not have advantage comparing to Tmall Genie, Xiaomi AI, and even Myna AI which are below CNY 400 (USD 60) and Rokid’s speaker is priced approximately at CNY 1,000 (USD 150). Considering the acoustic quality, Sonic’s SRS-XB10 is less than CNY 350 (USD 52) while Rokid has no accumulated reputation in this area.
Before the young generation become the dominant and maturity of the smart home, AI product and solutions for ToC business will still meet difficulties. ToB market is well-prepared to welcome new technology to help themselves become more efficient and competitive in the market, so the acceptance of AI technology is comparatively higher than ToC market.
5G and Security
AI functions in security area can be amplified by realizing 5G technology. As previously mentioned, 5G technology has the potential to reform industries and drive the development of other technologies.
The broadened bandwidth can improve current AI technology’s application in security. i) It can allow much more data transmission; ii) the cloud computation capability is improved; iii) the image processing and analysis’ time costs less; iv) reflection time is shortened; v) response and the reaction can be taken faster.
The scenario in a campus setting might like followings. A student can enter and exit campus using the face recognition technology and which might take less than a second; the student buys books or other things on campus using the same technology in checkout process instead of cards; if the camera detects some suspicious behaviors like people who wandering outside student forms with certain poses that are marked as suspicious, the security system will soon response and send feedback to security officer; other cases.
Other scenarios might have even more complicated circumstances for the security system to judge. The increasing demand for security management efficiency gives AI technology a playground to testify technology’s application capability. Security is only a vertical among all other AI application areas, others like healthcare, smart home, mobility, etc. are waiting to be changed by the new technology.
SenseTime and Megvii
SenseTime and Megvii are the two most mentioned AI company in security area due to their core competency in CV and landing capability. These two unicorns have entered a later state in financing and can be considered as the most successful AI startups. Venture capitals and investors believe in these companies’ potential and acknowledge what they have done for companies and even the AI industry’s development in past years.
SenseTime’s business model is 1+1+X, which is R&D + Solutions & Products + Industries and Verticals. In security, SenseTime cooperated with Huawei in providing smart city solutions to Chongqing, Hebei, Guangdong, etc. Besides the smart city construction, it offers security solutions to Boao Forum for Asia, The Summer Davos Forum, and other international conferences.
Megvii, previous Face++, is an AI startup started from CV technology (i.e. same as the technology supporting Snapchap’s camera effects). Its strengths in facial recognition enabled it to expand its application from providing technical support to effect camera apps on the smartphone to public security – a transition from entertainment to public service. Its solution has been used by government regulation and public service departments such as police department, custom, etc. Megvii upgraded its facial recognition AI platform to current Brain++ AI algorithm engine and changed the company’s name to current from Face++. The upgrade reveals its ambition to provide integrated solutions and services based on its AI technology instead of only on facial recognition.
In conclusion, AI is a foundational technology that could apply to numerous industries and verticals. At a time that AI is still an ambiguous concept for the mass, the ToC business for AI is not a perfect alternative for AI startups, who need positive cashflows to attract investors and increasing profits to support its running. The dilemma met by Rokid in its smart speaker can explain the ToC business' difficulty. ToB is easier than ToC is only because of the different incentive in pursuing the interest. To sustain a business, executives and managers always have the need to improve their business performance through a cost-efficient way and are willing to embrace the new technology and ideas.