Zhao Jinglei: ReadSense Ltd Pay more attention to the commercialization of the ground, research AI c

Although the focus of speech recognition and machine vision to identify companies behind closed doors, has not released sales data, but to determine the health of an enterprise, sales income statements is the most intuitive.

China is not a country with a slow, Chinese hard enough, most of the investment institutions are the pursuit of tuyere heat waves and fast growing enterprise, quit as soon as possible, artificial intelligence is no exception.

Calls in the first half of 2016, VR is highest, the second half of the VR is investors profit interrogation, another event occurs in the second half of the year is based on the deep learning of big data, visual recognition, speech recognition and artificial intelligence company hyped, needless to say, investors and the media of the patient can't more than 1 year, so come back in the Spring Festival of 2017, all people will be painting style, "your product profit?"

So, although the focus of speech recognition and visual companies behind closed doors, has not released sales data, but to determine the health of an enterprise, sales income statements is the most intuitive.

Shanghai jiaotong university, artificial intelligence, Dr Zhao Jinglei is trained, he

Is a focus on "deep learning" commercialization, and resist the hype deep learning technology CEO, he thinks, the AI technique is infrastructure, and industry can be transformed into productivity, research of AI company money is a bottomless pit.

Based on efficient management, based on consumer upgrades, visual identification with embedded speech recognition will become a kind of infrastructure to the details in the life and work. This is to follow the trend of development, for example, a camera on the size, cost, function, pixels are constantly optimized, then the camera still have space to grow up? Affirmation is some, no matter from the hardware and software will have an iterative update, add visual identification module, may be in a short period of time to have the function of camera price increases, but with a mature technology, to return to a reasonable price, everyone can consume.

Zhao Jinglei think that with the advent of the era of intelligence, people has higher requirement for artificial intelligence.

Previous computer vision in industry breakthrough is mainly focused on the algorithm of the cloud, the cloud scenarios computing resources is unlimited, can use a large cluster to run very complicated model. Recently Alphago defeated lee se-dol, hundreds of machines on the computing power is over a human being, in less than a year, the technology is not worth showing off.

For a new generation of intelligent machines, such as robots, VR/AR, cameras and other equipment with a camera, visual information processing have two typical characteristics, one is a real-time, one is localization. To solve the problems is to make intelligent machine eyes deep learning algorithms in the front, in an embedded environment very real time, low energy running. Read surface technology is mainly to solve the above problems.

Zhao Jinglei machine perception of environment can be divided into three levels, the first instinct layer, the second is the acquisition layer, the third is the professional, instinct layer is a child from was born with a visual ability, such as people's awareness and perception of space. Acquisition layer is the most people rely on the constantly learning to master the day after tomorrow, such as the recognition of text, and a variety of items. Is certain professional talents with professional layer, such as judging patients by medical images.

Zhao Jinglei think visual ability is the instinct layer every consumer and civilian camera needs to have the processing power, precipitation in the future should be based on the front end of the visual module, plug and play. Acquisition layer visual ability more in the future should serve public cloud precipitation. Layer and professional ability should serve vertical visual precipitation in the future.

Read surface technology products currently focused on the hardware visual module is the instinct layer, including "human perception module" and "spatial perception module". Read the following commercial reason is "basic module + from service". Zhao Jinglei demand for products is plug-and-play + one-stop solution, which means that the front vision module will do more and more light, and from the service will do more and more heavy. At the same time, surface technology or reading a very focus of enterprises, the current products to the market is still spending and civilian, and no package cover all scenarios.

In the Angle of product commercialization, Zhao Jinglei think points, according to the market can be divided into and incremental market and stock market in a traditional camera products are fit to load visual identification module, realize the function of intelligent ascension, such as the original family, the community of all kinds of cameras can be upgraded to have intelligent camera of human perception, this is the stock market intelligent upgrade.

Is another piece of commercialization, incremental market, all the new type of intelligent machines need the one-stop visual identification of plug-ins, such as a variety of camera equipment, for the present, Zhao Jinglei small statistics, the stock market in the present single volume accounted for more than, present clients are the IPC form camera such large enterprises as well as the traditional home appliance, incremental market is given priority to with all kinds of middle and small startups. He believes that in the future, the incremental market quickly.

It's not Zhao Jinglei first startup, in the mobile Internet startup period, he also made the APP, is clearly very love tossing and entrepreneurship, admits he do a CEO for many years, in the new era of struggle, there is still not good at, read surface headquarters in Shanghai, in nanjing and Beijing r&d division, in terms of sales together, has set up a branch in shenzhen, now he will worry the talent introduction.