Deepmind new AI system logic reasoning ability significance homefront

The new DeepMind AI system has the ability of logical reasoning significance of Tencent Francisco, according to foreign media reports, Alphabet’s artificial DeepMind Intelligent Company developed the latest AI technology can be used to address the need for external memory logic in order to complete the task, which marks the artificial intelligence technology has taken a big step forward. Artificial intelligence system, also known as neural network, it can identify the image, translation language or even go. However, it is limited in the complex relationship between the performance of the data or the various variants, so the artificial intelligence system is rarely used in the work of logical reasoning. DeepMind, 12, published a paper in science magazine this month, claiming that it has made great progress in this area, which uses external memory to improve the ability of artificial intelligence systems logic reasoning. The neural network is combined with the external memory, the neural network can not only learn, but also can use memory to store and recall some information, so as to clarify the complex relationship between data or various variants. This allows the neural network to solve more complex problems, such as in the London subway network navigation and unlock logic puzzles. Although solve this kind of problem and can give people a sense of bright eyes, because the programming algorithm can also do this, but this kind of hybrid system can in the absence of any default rules under the premise of completing this task, it makes sense. Computer scientist Joshua · of the University of Montreal in Canada (Yoshua Bengio); the GIO said, although this is not a new method, but the progress is this thesis made extraordinary. The memory magic neural network can learn new knowledge by strengthening the connection between components similar to the virtual neuron. If there is no memory, such a network may need to look at thousands of times in London subway network diagram in order to learn the best route planning. DeepMind called the new AI system "Computer Neural" (Differentiable), which can draw maps it has never seen before. It will be the first to use similar structure map, randomly generated (composed of representatives from the subway station and the subway line line) to train the neural network to learn how this process is stored in the external memory about these relationships and the description of the answer and their related problem. Meet the new map, AI DeepMind system will be a new relationship to memory, such as the London Underground network as an example, the new relationship is the line between different subway station, later planning line put it out. DeepMind’s AI system is also used to solve problems that require reasoning to solve. After 20 different types of question and answer training, it learns to accurately infer. For example, the AI system can be inferred from the "John picked up the football" and "John on the playground," the two sentences correctly". The accuracy of the AI system for answering such questions is more than 96%. In response to the need to push.相关的主题文章: