Wang Xiaogang, dean of Shangtang Research Institute, said that the accuracy of artificial intelligence (AI) face recognition technology in 2014 exceeded the accuracy of human eyes for the first time, but the current error rate can reach one billion. If the face recognition of the machine is equivalent to a 4-bit password, it is now equivalent to an 8-bit password. The accuracy of face recognition has increased by 4 orders of magnitude within 4 years. Wang Xiaogang, Dean of Shangtang Research Institute, gave a speech entitled “The Present and Future of AI Empowerment†at the MIT Technology Review Summit, explaining how artificial intelligence can change life and future through the cloud and the end. Wang Xiaogang recalls that in the past decade, the most outstanding representative of the rapid development of artificial intelligence is deep learning. There are three factors that drive deep learning: first, big data, second, cloud computing power, including GPU and AI chips, and third, deep learning algorithms. For example, Wang Xiaogang, in 2014, the face recognition of the machine can reach one ten thousandth of the error rate; now the error rate can reach one billionth. In other words, if the face recognition of the machine is equivalent to a 4-bit password, it is now equivalent to an 8-bit password, and the face recognition performance has been improved by 4 orders of magnitude. As the algorithm improves, the application boundary continues to expand, from the first 1:1 identity comparison to the later dynamic control, such as arresting suspects. It is now possible to search for faces from hundreds of millions of images throughout the city and restore human movements. This high accuracy actually benefits from the power of the neural network, by training more than 1,200 layers of neural networks, thus exceeding the ability of human eye recognition. The neural network has evolved from the 5th floor in 2012 to the current 1,200-layer process, and the complexity and depth of its network are constantly improving. But why use only a small network in the front-end application, but need to train a network of 1,200 layers in the background? Wang Xiaogang pointed out that the reason is that in order to get a small network with high recognition rate, you first need to get a very deep and very strong network to learn. He likened that this small but high recognition network is like a primary school student. The information is like a book. If you let a primary school student directly extract knowledge from a large number of books, it is very difficult to find a teacher. The teacher has a very strong learning ability. This is the 1,200-layer network. It can first extract knowledge from a vast amount of data, and then teach a smaller network through knowledge dissemination to achieve good recognition performance. The improvement of the algorithm can bring a lot of applications and make the city more secure. For example, if a criminal has been anonymously named more than ten years ago, even if he changed his ID card, he was eventually captured by the dynamic face control system. In addition, the Public Security Bureau used the face recognition system to combine tens of thousands of cameras in the city to successfully find the lost elderly. Or children. According to the accumulated results of Shang Tang in the past few years, in 2014, it was able to train the machine with 200,000 faces to achieve an accuracy rate of 98.5%, while the number of people was 97.5%. In 2015, the training was performed with 300,000 faces. 99.55% accuracy; in 2016, 60 million face training can achieve a million percent misunderstanding rate; in 2017, 2 billion face training can achieve a misunderstanding rate of one hundred million, such a The rate of misrecognition has been truly applied to all walks of life. On the hardware side, Wang Xiaogang also mentioned the strategic cooperation between Shang Tang and Qualcomm. He believes that if artificial intelligence wants to be popular, then AI will go to the front-end equipment. These front-end devices are inseparable from the chip. At present, Shangtang's face unlocking technology has been used by more than 100 million mobile phone users, and it must rely on Qualcomm's chip support.
Shenzhen WeiKa Technology Co.,Ltd. is a professional e-cigarette manufacturer and enterprise integrating R&D, design, production and sales and service of the electronic cigarette atomizer products, established in 2020.
We are actively developing and producing all kinds of reliable and unique atomizer products. We centre on research, developing, high quality and safety, and focus on better vaping experience in all aspects to our customers. We have advanced production equipment and strict production control system, each process has strict management, from production to delivery have been strictly tested to ensure the classic quality. Zgarvape for the future!
Aurora E-cigarette Cartridge is loved by the majority of consumers for its gorgeous and changeable colors, especially at night or in the dark. Fruits Flavors series, plant flavors series & special flavors series.Up to 16 different flavors for choices. A new design of gradient our disposable vape is impressive. Our vape pen and pods are matched with all the brands on the market. You can use other brand's vape pen with our vape pods. Aurora series Pods systems, the first choice for professional users!
16 Available Flavors (3Pods*2ml):
Amazing Cantaloupe / Coke on the Rocks / Ice Coconut / Ice Lolly / Iced Mung Bean / Longjing Tea / Mango Feast / Mountain Spring / Peach Tempting / Pineapple Juice / Refreshing Melon / Refreshing Mint / Rich Strawberry / Rose Litchi / The Ocean / Watermelon Juice
Aurora Pods 1.0,Pod System Vape,Pos Systems Touch Screen,Empty Disposable Vape Pod System,400Puffs Pod Vape System ZGAR INTERNATIONAL(HK)CO., LIMITED , https://www.zgarette.com
Aurora Disposable Vape Device Vaping Pod Systems
