Horizon CEO Yu Kai: edge computing is the cornerstone of autopilot.

Horizon CEO Yu Kai: edge computing is the cornerstone of autopilot.

Author: Quan Xiao Xing

With the development of the artificial intelligence industry and the spread of application scenarios, artificial intelligence has been rapidly integrated into people's life and production processes, and with the updating of artificial intelligence products and the diversification of Intelligent Computing scenarios, the requirements for artificial intelligence hardware, algorithms and chips have been improved, while artificial intelligence has been improved. The main direction of energy migration is also moving from cloud to edge.

It can be said that "edge computing" is becoming a major concern of the AI industry, from chips, terminals, to algorithms, applications, new technology trends for each link have posed a serious challenge.

September 19, 2018 World Congress of Artificial Intelligence "Edge Computing Intelligent Future" Summit Forum held in Shanghai. Horizon founder and CEO Yu Kai attended the Forum on the same day, saying, "AI makes the city safer and better, and computing requires a lot of sinking to the edge, must rely on reliable, efficient edge AI chips."

Yu Kai believes that the future development of AI will inevitably lead to the establishment of smart cities, and will completely change the traffic, driving, so that there are a lot of human-computer interaction in these industries, and proposed that in the AI for large-scale data explosion in urban life, edge computing will become an inevitable trend, "edge computing is so" The important thing is that it will really make life safer and better for the whole city.

"According to a report from a research institute, the global data volume will soar 10-fold to 163ZB by 2025, far more than any Internet company now has in reserve." Yu Kai said he thought it was necessary to process such a large amount of data in urban application scenarios, and to perform calculations in real-time, reliable and low-energy environments, many of which needed to be done at the edge.

Yu Kai quoted the three Kingdoms'romance "dividing time into time and dividing time". He believed that in the PC era, with the development of the Internet, it once evolved from edge computing to cloud computing. From the beginning of mobile Internet to the present Internet of Things, more computing moved from the center to the edge, especially in tunnels and tunnels. Underground garage and other vehicles without signal environment, or for security enterprises with high reliability requirements, if relying on cloud computing with the help of signal and network speed, the reliability will be greatly reduced, and affect the further popularization of artificial intelligence technology.

In recent years, some areas actively attract data centers, Yu Kai believes that data centers will generally be built in areas with less demand for electricity, which is due to the data center for the power supply system has a stronger demand, while the edge of the calculation needs to meet the requirements of low energy consumption, "it can be said that edge computing for processors. It's very demanding, and that's the cornerstone of the technology that's at the heart of our smart city today.

Yu Kai interpreted the more important "Moore's Law" in physics, which means that the number of transistors that can be accommodated in IC will double every 18 months and the performance will double. He thought that with the approaching of the atomic limit, the original Moore's Law can not meet the needs of the development of the artificial intelligence industry. By using scenarios, tasks, and software algorithms to drive the "New Moore's Law," and taking unmanned driving as an example, it is shown that the total amount of computation increases by an order of magnitude for each level of unmanned driving as the level rises, which also means that, if it is necessary to achieve all-weather, all-road conditions completely unmanned, it is reasonable. The amount of information that you need to achieve is 1000T.

"Basically by 2025, according to the new Moore's Law, the amount of power that a thousand dollars can buy is equivalent to 1,000 T, which is exactly the same as our human brain today, and we continue to develop matching software systems that will make the real five-level unmanned by 2030. It's possible to drive. " Yu Kai said.

Therefore, Yu Kai pointed out that with the popularity of large data brought about by the intervention of large computing, and the improvement of simulation performance, the edge of the AI processor will be the main battlefield of future scientific and technological competition, and will become an AI industry technology commanding heights.

In order to get the advantage of the first-mover in edge computing, it is not enough to rely on software alone. It is necessary to combine software with hardware deeply, and the importance of the combination of software and hardware is put forward. "There used to be only one company and only one person, Steve Jobs and the Apple company he led, so we saw the entire mobile age, the only company in the world that did the combination of hardware and software, the chip itself, the software operating system itself. "

At the same time, taking the unmanned driving scenarios using artificial intelligence technology as an example, from the scale of the automotive market, the United States is 17 million vehicles, China will maintain the size of 25 million vehicles this year, so China's advantage in the number of scenes, but also for the development of artificial intelligence industry and technological progress, bring a series of advantages.

In this regard, Yu Kai said, "(in the various scenarios of artificial intelligence) we think the highest or automatic driving. There will be gains along the way, including Smart City, Smart Retail, Smart Manufacturing, which are all applications of edge processors..."

Waonews is a news media from China, with hundreds of translations, rolling updates China News, hoping to get the likes of foreign netizens