During the Fifth Future Network Development Conference, Baidu CTO Wang Haifeng shared the new characteristics of AI technology in accelerating industrial practice. Wang Haifeng pointed out that artificial intelligence is currently in the stage of industrial production, and integrated innovation is becoming the new normal.
In Wang Haifeng’s view, in the new development stage, artificial intelligence technology is becoming more and more complex, and integration and innovation are becoming more and more abundant. Specifically, the current integration and innovation of artificial intelligence technology can be expounded from the integration of software and hardware, the integration of knowledge and deep learning technology, the integration of cross-modal multi-technology, and the integration of technology and scenes.
In terms of integration and innovation of software and hardware, hardware design and software algorithms are moving towards fusion. Take Baidu Kunlun, a general-purpose AI chip in the cloud, as an example. This chip is specifically optimized for AI technologies such as voice, natural language processing, and images, and effectively supports deep learning frameworks such as flying paddles.
In terms of the integration and innovation of knowledge and deep learning, constructing a multi-dimensional heterogeneous super-large-scale knowledge graph from massive data can raise the level of knowledge to 550 billion. At the same time, large-scale knowledge can also be integrated to achieve knowledge-enhanced in-depth semantic understanding, and continue to improve the effect of semantic understanding in natural language judgment, automatic question and answer, semantic matching, and sentiment analysis.
In terms of cross-modal multi-technology integration and innovation. Through the knowledge graph associating cross-modal information, and using natural language semantics to express the semantic space that combines different modalities such as language, vision, and speech, it can also achieve knowledge-enhanced cross-modal deep semantic understanding.
With the further deepening of integration and innovation, artificial intelligence technology is becoming more and more complex, and it is necessary to continuously lower the threshold to meet the rich and diverse industry applications. According to reports, as a deep learning open source open platform dedicated to lowering the threshold and meeting the needs of developers in different industries, at different stages, and at different levels. At present, Flying Paddle has gathered 3.2 million developers for collaborative innovation, and the number of enterprises and institutions it has served has reached 120,000.
Source: C114 Communication Network
Author: Wen Meng