SenseTime should adhere to its mission: to adhere to originality and let artificial intelligence lead human progress.
In the final time of 2021, SenseTime successfully listed in Hong Kong. This artificial intelligence company, from its inception to financing, to the listing, can be described as full of thorns.
As early as 2017, SenseTime set the company's mission: to adhere to originality and let AI lead human progress.
But then I came across a small episode while designing a new logo. After listening to SenseTime's entrepreneurial story, the designer made a suggestion: SenseTime should develop into an international hard technology company, and the Logo is best to only have a pattern or English, because it is difficult to associate high-tech when seeing Chinese words...
This isn't the first time SenseTime has encountered the obstacle of stereotypes, nor will it be the last.
4 years later, in 2021, SenseTime has become a new giant of artificial intelligence with annual revenue of billions of yuan, technology widely used in manufacturing, transportation, cultural tourism and other industries, flowers and applause more and more, although after submitting the prospectus, encountered some troubles, but fortunately in the final successful listing.
Since its inception, every road that SenseTime has traveled reflects the development level of a country's science and technology and the epitome of industrial competition in a big country. This artificial intelligence company, which has been pinned high hopes since its birth, but has been bumpy in the process, and has now a market value of more than 200 billion, has done right in the past 7 years?
"Artificial intelligence" is not a new word, but in the past few decades there has been no clear industrial development path, and even what exactly is "artificial intelligence" is also a matter of opinion. When SenseTime was founded in 2014, it was actually going to take a path that no one had ever taken.
The beginning of SenseTime technology is very "excellent student" characteristics: issuing papers and winning awards.
From 2011 to 2013, in the field of computer vision, there were a total of 29 articles on deep learning worldwide, 14 of which came from SenseTime's founding team.
In 2014, the Multimedia Laboratory of the University of Chinese in Hong Kong, led by Tang Xiaoou, released a face recognition algorithm called DeepID, with an accuracy rate of 98.52%, which greatly exceeded the human eye recognition ability (97.53%), breaking through the industrial application standard, and the multimedia laboratory of CUHK was also known as the "Whampoa Military Academy in the Field of Computer Vision". In the same year, Tang Xiaoou founded SenseTime and began to translate the research results into commercial applications.
In 2015, SenseTime participated in the ImageNet Massive Visual Identity Challenge, and with a new generation of algorithms, the accuracy rate exceeded Google's previous score by 6%, winning the championship and breaking a new record. The algorithm was then published on CVPR, kicking off the prelude to AI's industrial revolution.
But to turn from laboratory results to practical applications, the first hurdle is scale. Laboratories and competitions are short-term, local environments, and once they are commercially applied, they may be all-weather and multi-scenario, which is a huge challenge for talent reserves, computing power foundations, and algorithm iterative updates.
In order to accelerate development, SenseTime began financing. According to the prospectus, SenseTime completed a total of 12 rounds of financing before the listing, with a total financing amount of up to US$5.2 billion.
Behind the continuous financing is huge R&D expenditure and infrastructure construction.
According to the prospectus, in the first half of 2018, 2019, 2020 and 2021, SenseTime accumulated R&D expenses of 6.99 billion yuan. Among them, R&D investment in 2018-2020 accounted for 45.9%, 63.3% and 71.3% of the revenue of that year, respectively, and in the first half of 2021, R&D investment was 1.77 billion, even exceeding the scale of revenue.
This money, on the one hand, is invested in the talent reserve. SenseTime's more than 5,000 employees, about two-thirds of whom are scientists and engineers, have won more than 70 global competitions and published more than 600 top academic papers, and almost all of the top Chinese scientists in the field of deep learning are concentrated in SenseTime's offices.
The other aspect is the construction of infrastructure. Human learning needs to be read a lot, and the deep learning of machines also has to process massive amounts of data and identify the required information from it, so it is also necessary to build a special "school" and "library" for artificial intelligence. To this end, SenseTime has built a large artificial intelligence computing center (AIDC) in Lingang, Shanghai, which is expected to have a total capacity of 3740 Petaflops and can process 23600 years of video in 1 day, which is only part of the larger "SenseTime AI device" plan.
So far, SenseTime has built a leading theoretical system, talent team and infrastructure around artificial intelligence, but whether these technologies and patents can find enough application scenarios to bring cost reduction and efficiency to enough users, I am afraid that few people have a bottom in mind.
How can the utility of these technologies be verified? There is always only one most effective way: actual combat.
In the nearly decade since the concept of "artificial intelligence" has entered people's field of vision, a large number of related technologies have been mainly applied to some head scenarios, such as traffic roads in cities, production lines in factories, etc., because under the condition of limited resources, it is necessary to give priority to solving those high-frequency and common problems. However, there are thousands of long-tail demands in cities, businesses, and lives, which seem to be very small and do not necessarily occur frequently, but there is a huge value that has not yet been tapped.
Take urban governance, for example. As a global metropolis, Shanghai manages a population of 30 million per day, involving thousands of scenes and services, garbage exposure, bicycle parking, green space maintenance, building fires, etc. are all subtle and important problems. However, the traditional management method can often only "see" the image but "can't see" the problem, and still requires a large number of manual meshers to strip to the scene to inspect. Society has to bear a huge human expenditure.
Urban governance is a huge and complex "giant system". To solve the massive problems, it is necessary to match a large number of AI algorithms, but the production of a single AI algorithm takes dozens of engineers for months. Therefore, often a small part of the needs can be solved through AI, a large number of daily scenarios, long-tail application scenarios have not been covered, artificial intelligence is inevitably regarded as a "sample project".
In response to this problem, SenseTime technology has given a new answer - SenseCore SenseTime AI device. With the blessing of the computing power layer, the platform layer, and the algorithm layer, artificial intelligence has the possibility of landing.
For example, Jiangsu Road Street in Changning District, Shanghai, has taken the lead in piloting AI+ network management since 2020, based on SenseTime's AI large device, mass-producing algorithm models for various long-tail scenarios, deployed to the "intelligent patrol screen" through the Ark City Open Platform, thereby transforming 1,000 cameras in the region into intelligent perception neurons, discovering urban diseases in seconds, and pushing them to managers for the next step in a timely manner.
With the support of SenseTime's "Ark City Open Platform", the daily work of Shanghai grid members has now changed from the "human sea battle" of constantly patrolling the streets to the interaction between "people" and "machines". It will gradually expand to more scenarios such as illegal parking of motor vehicles and water accumulation on roads, so that the urban management system can operate more efficiently, safely and orderly. Shanghai's urban management level has also been repeatedly praised by netizens in recent years.
It can solve the complex problems of urban management through "smart city", and naturally can solve more relatively simple application problems. From the virtual and real navigation guide application by the West Lake in Hangzhou, the AR cultural and creative calendar of the Forbidden City in Beijing, the AR postcard of China Post, to the intelligent driving system of electric vehicles, the maintenance of the power system along the high-speed rail, and the automatic detection in the automobile manufacturing process, the landing application of SenseTime has shown a trend of fire, and it is in the four major fields of smart business, smart city, smart life and smart car.
In terms of smart business, SenseTime has cooperated with enterprises to build the "SenseTime Ark Enterprise Open Platform" to digitize and automate its workflow, which has been deployed at about 6,000 customer sites. For example, after applying the enterprise Ark, China FAW Group can automatically detect more than 34 defects in metal parts in 6-12 seconds, with a detection rate of more than 99%, and can operate around the clock with high stability, which greatly improves production efficiency.
In terms of smart cities, SenseTime's Ark City Open Platform contains more than 14,000 artificial intelligence models, which can be seamlessly integrated with the city's IT infrastructure, improving the safety, efficiency and convenience of the city. As of June 30, 2021, City Ark has been deployed in 119 cities at home and abroad.
In terms of smart living, as of June 30, 2021, SenseME and SenseMARS software platforms have cumulatively supported more than 450 million mobile phones and more than 200 mobile applications, including mobile applications with more than 500 million monthly active users; the SenseCare software platform provides a variety of artificial intelligence tools to empower medical services such as diagnosis, treatment and rehabilitation. Five SenseCare modules have been certified by the NmPA and two CE certifications.
In terms of smart cars, at present, SenseAuto SenseTime Smart Car Platform has accumulated more than 1,100 intelligent car technology patent assets, cooperated with more than 30 domestic and foreign head partners, and covered a total of more than 20 million vehicles in fixed-point mass production projects.
On the basis of ai devices, SenseTime's four major businesses have flourished. The advantage of verifying the practicality of technology through a large number of practical applications is that on the one hand, a wide range of commercial applications will produce a large amount of data, which can be used as the basis for subsequent iterative improvement; on the other hand, with the technical reserves, from the vision to a vivid business case and real revenue, the confidence of enterprises on the established development path has been continuously enhanced.
A technology company, founded only 7 years ago, has been commercializing everywhere, sparks. How exactly did this come about?
The spirit of a company depends largely on the character of the founding team. SenseTime's unique model of making artificial intelligence blossom in multiple industries actually comes from the founding team's insistence on the company's positioning.
Time back to May 2017, SenseTime held its first executive meeting, and one of the longest-consuming topics was about the company's mission and values, and before that, the definition and construction of corporate culture had actually been brewing and preparing for half a year.
At the meeting, SenseTime co-founder, chairman and CEO Xu Li proposed that SenseTime's mission is to "let AI lead human progress", and founder Tang Xiaoou added four words before this sentence, "adhere to originality".
At that time, artificial intelligence was just entering the early stage of industrialization, and the industry saw more of the market potential of AI technology in solving local areas such as driving, traffic management, and face recognition. But with its insistence and understanding of original technology, SenseTime has a longer-term vision, so it can take a grand goal of "letting AI lead human progress" as its mission. And "let AI empower hundreds of industries" is an important way to achieve this mission.
In other words, SenseTime wants to be an "all-rounder" of deep learning, making itself like a power plant, where AI is needed, it can deliver AI capabilities to wherever it is needed.
It is under this long-term positioning that SenseTime has launched a basic platform with SenseCore AI device as the core, and is committed to creating "two tools" through innovative AI technology.
The two tools refer to the use of AI as a new productivity tool and an interactive tool. On the one hand, the AI large device has greatly reduced the production cost of AI models, thus deriving senseTime's commercial landing ability of "AI-enabled hundred industries", reducing costs and increasing efficiency for traditional industries, and becoming a "productivity tool" for a large number of enterprises. On the other hand, SenseMars platform contains 3500 artificial intelligence models, supporting perceptual intelligence and mixed and augmented reality systems, which also plays an important supporting role in creating new interactive scenes such as metaverse.
SenseTime takes the AI device as the cornerstone to support the research and development of the underlying technology, and at the same time grasps the two core scenarios of productivity and interaction, extends the AI capabilities to all walks of life, covering many fields and scenarios in urban management, enterprise services and personal life, and becomes a hard technology company that can empower the development of the industry externally.
end
After experiencing the challenge of "seeing Chinese characters is difficult to associate with high technology", a new logo containing Chinese characters, graphics, and English was finally successfully designed and put into use.
At the time of activation, SenseTime explained the meaning of the Logo on the public account, using a very abstract paragraph:
"The infinitely long curve is the abstraction of the universe, one end is connected to the infinite past, the other is connected to the infinite future, there are only random undulations of irregular and inanimate in the middle, and the peaks of the high and low are like grains of sand of different sizes, and the whole curve is like a one-dimensional desert formed by all the sand and gravel in rows, which is unpredictable. You can walk infinite distances forward and backward along it..."
This passage is excerpted from "The Three-Body Problem". Just as before the novel came out, almost no one would have connected the Chinese with the top science fiction. SenseTime has gone from winning awards in the lab to applying technology to a large number of scenarios in the past 7 years, and if you put aside fashionable words such as artificial intelligence deep learning, it is not difficult to find that the background of this company is still the diligence and wisdom of the Chinese nation.
At senseTime's launch ceremony, CEO Xu Li said: "Enterprises in each era carry different hopes and responsibilities. The inheritance of China's entrepreneurial spirit from generation to generation has jointly spawned the magnificent scene of the past for more than 40 years, and has pushed our proposition from survival and catching up to development and pioneering. A colleague once wrote a letter at the child's coming-of-age ceremony, "When we struggle to this point and no longer suffer from food and clothing, all that is left for you is to change the world, even if it is only a little bit'. What I want to leave to SenseTime is our mission of 'insisting on originality and letting artificial intelligence lead human progress'. ”