Global Artificial Intelligence Breakthrough Technology TOP10 in 2018

Artificial intelligence is a master of high technology, wide field, multi-dimension and interdisciplinary. From the pure computer application based on big data and around the Internet, it gradually derives to all aspects of people's daily production and life, and improves and changes us in subtle places. At present, many new technologies and models have been put into practical use gradually, but most fields are still in the stage of popularization, experiment and research. How to grasp the great opportunity of popularizing artificial intelligence technology and let more people use artificial intelligence just like popularizing mobile phones is what our generation must face. The development of the times "must answer questions".

In 2018, AI technology has made breakthroughs in many aspects. Domestic and foreign technology companies are constantly trying to apply AI to more fields. Whether they are technology giants or start-ups, they are committed to continuous innovation and promote technological progress. Next, let's look at ten outstanding AI fields at home and abroad. Breaking technology.

1. Machine Translation Based on Neural Network

Reasons for selection: Translation is the most important branch of natural language processing and also a relatively difficult branch. In the early years, machine translation was also regarded as "low-level translation" and was ridiculed. Nowadays, the accuracy of machine translation of neural network has been greatly improved, which is comparable to that of professional manual translation. Google Translation, Microsoft Voice Translation and Sogou Speech Recognition are all based on this technology.

Technological breakthroughs: Machine translation has been a research field for decades. Its technical core is a deep neural network with massive nodes, which can automatically learn translation knowledge from corpus.

The process of human brain processing language is undoubtedly one of the most complex cognitive processes. Many people once believed that machine translation could not reach the level of human translation. Multilayer perception in neural networks attempts to simulate the process of multilayer transmission of human brain nerves, but usually does not exceed three layers. In 2006, scientists proposed a deep learning algorithm for neural networks, which makes it possible to train neural networks with at least seven layers. Because it can simulate the process of multi-layer depth transmission of human brain neurons, it has a very obvious breakthrough performance in solving some complex problems.

In March this year, Microsoft announced that its machine translation system had reached the professional level of human beings for the first time in the translation of general news from Chinese to English, and achieved another milestone in natural language processing. It had advanced machine translation beyond the time of amateur human translators by seven years.

35d2122ba10823d22aee5cd6e7b393e81545aa44

Significance: Machine translation based on neural network directly improves the effect of word-by-word literal translation and upgrades to the translation of the whole sentence as a unit.

Comments: Machine translation is not a problem in the world.

2. Robot autonomous navigation based on multi-sensor cross-border fusion

Reasons for selection: The ultimate goal of robots is to provide intelligent services for human beings. Autonomous navigation is a technical barrier that human beings have been trying to overcome in recent years. PowerVision can realize autonomous navigation and intelligent recognition for underwater robots through the fusion of sonar and vision sensors, and in the field of intelligent robots. Breakthrough progress has been made in China.

Technological breakthroughs: As the application scenarios and tasks of robots become more and more complex, it is difficult for a single sensor to meet the application requirements. Multi-sensor information fusion is highly dependent on hardware resources. Based on the core technologies of sensors, inertial navigation, motion control, cameras, visual detection/recognition, SLAM and in-depth learning accumulated by PowerVision for many years in the robot industry, through embedded research. The system architecture and optimization design of the integrated platform breaks through the limitation of hardware resources of the mobile platform, makes the underwater vehicle more accurate, intelligent and comprehensive perception of the target, and has the ability to lock, detect, identify and follow the underwater target.

550f38c79e4e29cc006a7ffcd8737160ef764db1

Significance: Through the combination of navigation control, sonar, visual technology and in-depth learning, robot positioning and autonomous navigation in the global environment, as well as intelligent applications, provide a strong technical guarantee for human exploration of broader underwater applications.

Comment: This may be the best way to land AI in the future

3. DuerOS Dialogue Artificial Intelligence System

Reasons for selection: DuerOS 3.0 can bring users epoch-making natural dialogue interaction, including the ability of emotional voice broadcasting, voiceprint recognition and other natural language interaction technology upgrade.

Technological breakthrough: DuerOS is a dialogic AI operating system developed by Baidu's Department of secrets. It has more than 250 skills of 10 major categories. DuerOS includes a complete interactive process from speech recognition to voice broadcasting to screen display, as well as the core technologies supporting interactive natural language understanding, dialogue state control, natural language generation, search and so on. These technologies support the implementation of application layer and capability layer.

On July 4, 2018, the latest DeuerOS 3.0 was officially released, which enabled the enabled product to realize voice multi-round error correction, complex progressive intention recognition and conditional intention recognition with logic, so as to more accurately judge the user's intention, and ultimately achieve functional dimensionality upgrade, using extended features to understand user behavior. Based on this, DuerOS 3.0 provides more than 20 cross-scenario, cross-device solutions, including screen device solutions, Bluetooth device solutions and industry solutions.