Yu Kai, CEO of Horizon Robot, shared a new market environment inspired by technological innovation from the perspective of a technical expert at the Global New Energy Vehicle Conference (GNEV7). Yu Kai was responsible for the research and development of artificial intelligence in Baidu from 2012 to 2015, including Baidu brain, voice search, image, advertising and search sorting and automatic driving in the cloud based on big data, large computing artificial intelligence, from July 2015 Creating the horizon and defining the horizon are the leaders of embedded artificial intelligence.
Yu Kai, CEO of Horizon Robot
What is embedded artificial intelligence?
Yu Kai introduced in detail that part of it is to realize artificial intelligence in the cloud through large-scale computing in the data center, but in local computing, in the case of non-networking, real-time environment awareness, man-machine Interaction, decision control. Especially in the case of autopilot, if a child suddenly crosses the road and sends a signal to the cloud to make a decision, this may be catastrophic. What if the network conditions are unstable? Therefore, under such a scenario, in fact, embedded artificial intelligence is a completely low-power localization calculation, which is actually a necessity.
This embedded artificial intelligence solution is also needed in many scenarios. Horizon is the leader in embedded artificial intelligence. This includes redefining how to implement low-power, high-performance computing from software, hardware, and even processor architecture. The main application scenario is undoubtedly the first one is autonomous driving; the second is called smart live.
Where is the next industry opportunity that can influence world change? Yu Kai believes that autonomous driving should be one of them. He reported from McKinsey and Bloomberg that in 2030, four out of every ten cars had a certain degree of autonomous driving, which of course included the possibility of assisted driving, and even 2030 should be achievable. Fully automatic driving at high speeds and even non-high speed automatic driving.
He explained that because we have a huge market, the Chinese auto market is already the largest in the world. Another embarrassing data is that China is also the highest in the world in terms of the death rate of traffic accidents per 100,000 people worldwide. How to make traffic travel is safer first and then more convenient, and then more interesting.
How to redefine traffic in the future without driver driving?
Internet companies, car manufacturers, and traditional suppliers must embrace this trend and find their place in their own industries. Secondly, in terms of user experience, the core content is that we will always underestimate the pursuit of laziness by human beings as a group. In the human-computer interaction industry of computers, it is more and more convenient to touch from keyboard to mouse to today. Let us be more and more The more lazy, the human-computer interaction can be done, and the endless pursuit of user experience will drive industrial change.
According to reports, during the period from September 2014 to November 2015, Google’s driverless driver had to intervene to drive unmanned driving. Otherwise, 13 collisions would occur, which occurred during the 420,000-mile driving. 341 man-made interventions, so it's actually really unmanned. Even in companies like Google, they can't actually be unmanned. The sensors and computing devices on the car add up to several cars, and even under such circumstances, there are a lot of risks, including stability and reliability.
Regarding enhanced learning, this is not only a problem of perception, but also a problem of decision-making, continuous decision-making, optimizing a long-term goal. Recently, the whole industry including Google and the horizon are studying the decision-making of autonomous driving based on enhanced learning. System, there may be a black swan event in this field. The other is calculation: calculating Moore's Law and redesigning the architecture.
In addition, Tesla announced that all new cars will be equipped with eight camera heads next year, all of which are high-definition. There are NVIDIA and PS2 computing platforms below, as well as 12 ultrasonic sensors, including the former image of millimeter wave radar. . However, their autopilot technology is not ready, so a strategy has been adopted: the technology is still not good, the hardware and sensors are installed first, and the technology is not moving data first. This is a new thinking and methodology. It will launch new cars in the next year, such as 40,000 cars, which will accumulate two kinds of data through the actual data of the vehicle. The first is that the road conditions are real-time data of the environment. The second type of data is how the driver operates and makes decisions under such circumstances and road conditions. The decision data of a driver who can accumulate almost one billion miles is behind a deep neural network that is constantly learning in the cloud. This is different from our past thinking. It is completely independent of machine learning. By 2018 they will have the ability to forward facing radar, which is a big data-driven autopilot that has the potential to greatly accelerate the development of the entire autonomous drive.
The raTIonal end to end autopilot decision developed inside the horizon through a large amount of data and computer simulation, through enhanced learning to learn the optimal path planning and decision-making algorithms, there is also a very important thing that we need to put artificial intelligence Technology is combined with the long-standing functional safety standards of the automotive industry, which is also an important issue.
In response to the current autonomous driving technology, Yu Kai concluded two points:
1. There are actually three types of participants in autonomous driving. One is the strong ecology of traditional OEMs and suppliers. This is the mainstream now; the first is Internet companies such as Baidu, Google, Uber; Classes are artificial intelligence technology companies like Horizon. We believe that traditional OEMs and suppliers and artificial intelligence technology companies need to respect each other. On the one hand, artificial intelligence technology needs to respect the safety understanding of the automotive industry. On the other hand, traditional OEMs and suppliers must rapidly develop the current artificial intelligence technology. It is possible to subvert the automatic driving technology like subversive Go, so they need to respect each other and embrace each other.
2, the impact of new technology on the world is always overestimating it in the short term, we will think that your imagination is not as fast as I imagined, so good. But in the ten-year dimension we tend to underestimate. Just like you couldn't think of how mobile Internet changed your life ten years ago, after ten years, it is often difficult to imagine that autonomous driving changed our lives at that time.
Tinned Copper Clad Copper TCCC
Corrosion-Resistant Copper-Clad Tinned Wire,Copper-Clad Copper Tinned Wire Production,Copper-Clad Copper Tinned Wire Processing,Copper-Clad Tinned Wire
changzhou yuzisenhan electronic co.,ltd , https://www.yzshelectronics.com