analysis- ai mind exactly what are you thinking-（分析- ai介意你在想什么）.docVIP

Analysis: AI mind exactly what are you thinking? (Research papers Download News) Today, the use of neural network is very extensive - but we do not know exactly how they work. Now, computer scientists finally began to glimpse AI ‘thought.’ Understand someone’s mind the idea will better understand his behavior. Artificial intelligence (AI) is also true. Recently, a new technology at the time of the neural network deal with the problem, ‘Take a Snapshot,’ will help us to understand exactly how neural networks work, and can bring us a better, more trustworthy AI. Over the past few years, deep learning algorithm to construct neural network (multilayer interconnect artificial neural) above in many areas of AI have made a breakthrough, including natural language processing, image recognition, medical diagnosis and chess. But there is a problem plaguing people that we do not know how they do that. Israel Institute of Technology, said Nir Ben Zrihem, deep learning system is a black box. ‘If it works well, that’s fine. If it does not work, you’ll be in trouble.’ Neural network is far greater than the sum of its parts. Their constituent units is very simple - is artificial nerves. Zrihem said: ‘You can not, pointing to an area of the neural network, and then said, this intelligence there.’ But the complexity of the connection means that we may have a learning algorithm to track the depth of the steps given result is reached. In this case, the performance of the machine is very like a prophet, but its results are not clear and certificate. To solve this problem, Zrihem and his colleagues created a deep learning of real-time image. They said the technology like the computer fMRI (functional magnetic resonance imaging), can capture the behavior of the algorithm to solve the problem. These images allow researchers to track the neural network of the different stages of the process, including the dead end. To obtain these images, the team to a neural