A week of international scientific and technological information (November 17-23) 【Full Text】

Science and Technology News This article summarizes international scientific and technological information for the week of November 17-23, 2018 for readers. Let's take a look with the editor, what happened this week:

artificial intelligence

Artificial intelligence chips based on memory computing technology were introduced, and performance has been greatly improved. Researchers at Princeton University in the United States have created a new type of computer chip focused on artificial intelligence systems. This chip overcomes the bottleneck of the processor's need to spend a lot of time and energy to obtain data from memory by directly performing calculations in memory. The chip uses a standard programming language, which can be applied to mobile phones, watches or other devices that rely on high-performance computing and have limited battery life. Laboratory tests show that the performance of the new chip is tens to hundreds of times faster than similar chips. The researchers said that the chip integrates the memory circuit into the programmable processor architecture. "If the previous chip is a powerful engine, the new chip will be a complete vehicle."

International gymnastics events will introduce Japan Fujitsu AI scoring assistance system. According to a report by NHK TV in Japan on November 21, the International Gymnastics Federation decided to introduce the AI ​​scoring assist system developed by Fujitsu Corporation of Japan. At present, the technical difficulty of the gymnastics competition has increased year after year, and the referee's judgment score has become more difficult. The system projects infrared rays onto the player's body and its surroundings to track the player's movements, and converts them into three-dimensional stereoscopic images in real time. Based on the images, AI analyzes the rotation and twisting of the body, combines the past performance data, and abides by the scoring criteria to judge the completion of the technology. The system is scheduled to be applied in the competitions held by the International Gymnastics Federation starting from the World Championships in October next year. In the actual game, the referee will combine the analysis of the AI ​​auxiliary system and give the player's score after comprehensive judgment.

CariGAN AI intelligently creates cartoon characters. Stanford University, City University of Hong Kong and Microsoft have jointly developed a system for drawing cartoons through a neural network. The AI ​​system consists of two sets of adversarial neural networks (GAN), called CariGAN. One set of neural network CariGeoGan is responsible for determining the facial geometry in a photo and mapping it to a comic model, and the other set of neural network CariStyGAN is responsible for handling style problems or applying artistic effects to geometric pictures. To enable CariGAN to make a relatively tedious picture vivid and interesting, the system was trained using thousands of hand-drawn images. CariGAN AI can also create comics by parsing the frames, and can reverse the design of the comics to determine what the characters look like. The researchers said: "We believe this may be useful for face recognition in comics."

The Style2Paints tool website uses AI to help colorize manuscripts. Recently, the Japanese tool website Style2Paints, which uses AI technology to assist in coloring manuscripts, ushered in the fourth generation update. After the user uploads the uncolored line art picture to the website, Style2Paints can be interpreted by AI. After a few simple steps, the line art is automatically filled with the corresponding color, and even effects such as shadows and light can be achieved. The special thing about Style2Paints is that its AI operation logic is to simulate people's real painting process and complete the coloring work step by step. In order to facilitate creators for subsequent creation, the downloaded file is set to PSD format. At present, the application of AI in the field of images has attracted more and more attention. Sensei AI launched by Adobe has brought many practical functions to software such as Photoshop.

robot

Stanford creates artificial gloves to increase tactile perception for robots. Recently, researchers at Stanford University made a pair of induction gloves for the robot, so that the robot can obtain the perception ability of imitating human skin and feel the difference of heat, pressure and other stimuli of various objects. The gloves are equipped with unique components that allow the robot to gain "tactile perception" capabilities when approaching moving objects. The researchers said: "One day, this technology will allow robots to acquire the same perception capabilities as human skin. For example, a robotic hand can touch a water basin, but it will not cause damage to it. Of course, there is still a long way to go to achieve this goal. go". The research was published this week in the journal "Science Robotics" and the paper details the working principle of the robotic glove.

Nvidia showcases the new Jetson platform, laying out the distribution robot market in China. At the NVIDIA GPU Technology Conference on November 20-22, NVIDIA introduced the new progress of the NVIDIA Jetson AGX Xavier platform specifically set up for robots, and announced that the Chinese e-commerce company Jingdong and Meituan have selected this platform to build the next generation of autonomous distribution robots. . The next generation of distribution robots needs to achieve large-scale computing performance in small packages. They need to recognize and react to pedestrians, cars, traffic lights, signs and other objects in real time. Jetson AGX Xavier is designed for such scenarios. Nvidia describes it as an AI computer suitable for autonomous robots. Although it is only the size of the palm, it can provide powerful workstation-like processing power. In addition, the Jetson AGX Xavier platform is equipped with a complete AI development software solution NVIDIA JetPack SDK to simplify the process and accelerate development.

Drone

University of Maryland researchers tested the delivery of kidneys with drones. According to TechCrunch, University of Maryland researchers carried out a test flight project to deliver kidneys in drones equipped with refrigerators. The researchers improved a DJI M600 drone to allow it to carry a refrigerator, and designed a wireless biosensor to monitor the status of the officer during flight. Before and after the drone delivery, the research team performed a biopsy on the kidney. Despite the threat of air cooling and the potential heat of the drone engine, the temperature of the refrigerator remains at around 2.5 degrees Celsius. The vibration and maneuvering of the drone did not seem to cause any damage to the kidneys. Although this mode of transportation cannot be practically used in a short time, after the risk is fully quantified and evaluated, kidneys, liver, blood and important medical supplies are likely to be transported in this way after a disaster.

VR / AR

Google gets a wheeled electric shoe that can enhance the VR experience. Recently, the US Trademark Office (USPTO) approved the "mobile shoes" submitted by Google. The device is oriented to virtual reality applications, allowing users to perceive walking while remaining in the center of the game area. All along, the big problem facing VR applications is that it is difficult for users to move around at will. If they want to walk in the open environment of VR applications, users need to occupy a wide physical space or use a bulky "omnidirectional treadmill". Now, the electric shoes applied by Google can solve this problem. When the user walks in the direction corresponding to VR, the small wheel will gently push the user back to the center of the game area, and the space requirement is greatly reduced.