From April 18 to 21, the 2021 Boao Forum for Asia Annual Conference was held in Boao, Hainan. The theme of this year's annual conference was "The Great Changes in the World: Together in the Great Global Governance Ensemble One Belt One Road Strong Voice".
On the afternoon of the 19th, Baidu CTO Wang Haifeng was invited to participate in a roundtable forum with the theme of "Artificial Intelligence in the Post-epidemic Era" at the Boao Annual Conference.
With the father of public key encryption technology and Turing Award winner Whitfield Diffie, AstraZeneca Chairman Leif Johnson and many other experts and business representatives, we will discuss the development status and future trends of the artificial intelligence industry.
Wang Haifeng: AI industry is seeing more and more integrated innovations
Wang Haifeng shared his views on "integrated innovation" in the forum.
On the one hand, more and more technological "fusion innovations" have emerged. Artificial intelligence is now a comprehensive function of computing power, algorithms, and data. This is very different from the earlier research on algorithms.
On the other hand, technology and industry are also undergoing deeper and deeper integration.
Artificial intelligence is infiltrating all walks of life, and each industry has its own problems to be solved, there are application scenarios to break through, and in-depth integration of technology and scenarios is required.
The "integrated innovation" level of technology is the integration of deep learning platforms and AI chips.
Take Baidu as an example. Flying Paddle is a deep learning platform. "But we are not only doing a deep learning algorithm platform, but combining it with a lower-level chip. With Baidu Kunlun chip and the Flying Paddle deep learning platform Joint optimization has greatly improved the effect. At the same time, the algorithm must be innovated according to the scene and data integration to achieve the best results." Wang Haifeng said on the spot.
The second is the combination of deep learning and knowledge, as well as multi-technology integration and innovation.
Baidu has constructed a large-scale knowledge graph with 550 billion knowledge. Through the integration of large-scale knowledge, it develops deep learning methods for knowledge enhancement. In solving the problem of semantic understanding, the effect of semantic understanding can be greatly improved under the same parameter scale.
At the same time, based on the association and unified representation of cross-modal information based on knowledge and semantic representation, on the basis of knowledge-enhanced semantic understanding, we have achieved cross-modal semantic understanding of speech, vision and language.
Wang Haifeng said, "The development of artificial intelligence technology today has become more and more complicated from the perspective of scientific research, but the threshold is actually lowering from the perspective of application."
When artificial intelligence began to be applied to all walks of life, not every industry had enough experts proficient in artificial intelligence algorithms.
Therefore, we need a convenient and easy-to-use platform that allows developers to focus on application development and accelerate industrial innovation.
For example, the flying paddle platform developed by Baidu solves basic development, training, deployment, model libraries, development kits, and other issues. It is open source and open, so that developers do not need everyone to write the first line of algorithm code and can directly call it. .
Significantly lower the threshold of applications, and accelerate industrial intelligence.
At present, Baidu Flying Paddle has gathered over 2.65 million developers, served 100,000 companies, and created more than 340,000 models based on the Flying Paddle platform. There are flying paddles in many fields related to the national economy and the people’s livelihood, such as cities, industry, electric power, and communications. Play a role.
In terms of AI security, Wang Haifeng said that AI's data security, model security, and the prevention of deep forgery and other security technologies are very important, which requires the joint efforts of all parties to build.
The sudden new crown epidemic has accelerated the landing of artificial intelligence in all fields of economy and society.
The new generation of artificial intelligence has become a key research area of my country's cutting-edge science and technology, and other countries in the world generally regard artificial intelligence as the core driving force of the digital economy in the post-epidemic era.
The guests of the forum shared and discussed the progress of artificial intelligence in technology and industrial applications from the perspectives of technological innovation, open source and data security around the future development trend and prospects of artificial intelligence in the post-epidemic era, and highlights frequently appeared.