Baidu’s Robin Li Introduces Firm’s Feedback-Driven Innovation

At the 2022 Baidu Create event on Tuesday, firm CEO Robin Li delivered a speech about feedback-driven innovation and its influence on China’s real economy, showing the tech giant’s expectations for the future golden decade of artificial intelligence development.

In the past year, there has been less discussion about “promising opportunity” and more attention given to the health of certain growth indicators.

From a graph of global GDP per capita in the past 2,000 years, it can be seen that in the first 1,800 years, the change was not obvious. In the last 250 years or so, it has increased sharply, driven by several scientific and technological revolutions. It was technological innovations that have driven great growth.

Robin Li pointed out that the symbol of the fourth scientific and technological revolution is the deep learning algorithm that would bring efficiency improvement and economic growth. Major innovations related to deep learning, including automated driving and intelligent dispatching systems in the fields of hydropower and energy, will have a great social impact like inventions such as automobiles and the internet.

Baidu’s Feedback-Driven Innovation

Robin Li gave priority to discussing “feedback-driven innovation.” Some scientists have done a thought experiment – how long will it take a blind man to solve a Rubik’s Cube if he rotates it once every second? The answer is 13.7 billion years. But if every time the blind man turns the Rubik’s Cube, someone gives him feedback, telling him whether he is closer to the goal or farther away from the goal, he would take two and a half minutes to finish the mission. This is the magical power of feedback.

Therefore, innovation can only be realized when we have the opportunity to enter the market and get feedback from users and customers.

Baidu has a lot of practical experience in “feedback-driven innovation” in its business development.

For example, Baidu’s Kunlun artificial intelligence chips are very advanced because they have been optimized for Baidu’s search service for 10 years. Baidu’s search service responds to billions of real users’ needs every day, and conducts 1 trillion deep semantic reasoning and matching every day, which can provide the truest and most timely feedback, thus forcing the optimization of deep learning frameworks and chips.

Baidu has revealed orders from Apollo Go, its autonomous driving services platform, every quarter since a year ago. The large order volume means that it can get the most useful feedback. Baidu noticed from social platforms that web users in many places hope that the unmanned vehicle travel service can be more accessible, which means that people’s acceptance of automated driving is very high, and commercial operation may be faster than expected.

Baidu’s business scope covers chips, frameworks, models, and applications. From the high-end chip Kunlun, to the deep learning framework “PaddlePaddle,” then to the large model “Wenxin,” there are key self-developed technologies that can generate various feedback at all levels.

Baidu’s Influence on the Real Economy

First, the threshold for developing AI technology is lowering.

PaddlePaddle is an open source deep learning framework developed by Baidu. At present, it has gathered 5.35 million developers, served 200,000 enterprises and institutions, and created 670,000 models, which has built a prosperous deep learning ecology. It enables developers to build AI applications like building blocks. Software is necessary for innovation.

As the technical base of many applications, the big model can effectively integrate natural language processing, computer vision, intelligent voice, and more, and can be optimized in combination with scenarios of various industries and businesses, thus getting rid of the traditional fragmentation development mode.

Second, AI empowers the development of the real economy.

Robin Li regards intelligent transportation as an intelligent dispatching system that can control traffic flow through intelligent traffic lights, thus improving efficiency. Through the intelligent transformation of the transportation network, traffic efficiency can be improved by 15% to 30%. Baidu’s intelligent transportation scheme has been implemented in 63 cities. Li predicts that the intelligent transportation scheme can make China’s first-tier cities no longer need to restrict vehicle purchases before 2027, thus activating automobile use. Before 2032, the congestion problem can basically be solved by improving traffic efficiency.

At the Baidu Create event last year, Robin Li said, “With the continuous lowering of the threshold of technology application, creators will usher in the golden decade of artificial intelligence.” But the process of realization will not be easy. This is the characteristic of technological development.

Opportunities and Challenges During Development

In January 2000, Robin Li returned to China to start a business. Just two months later, the internet industry shook and the world’s market value evaporated by 8 trillion yuan ($1,184 trillion). At that time, many people chose to give up and those who persisted made the golden decade of internet development.

Ups and downs like this will still happen one day in the AI field. The silver lining is that in the past year, AI has changed its direction both at the technical level and at the commercial application level.

At the technical level, AI has moved from understanding content to automatically generating content such as pictures, texts and videos.

The most representative change in business applications is automated driving. In the past, everyone thought that L2 through L5 came step by step. But Robin Li believes that after L2 comes L4. Because it is difficult to define the accident liability of L3, it takes longer to popularize it. As of September last year, Baidu’s L4 automated driving service orders exceeded 1.4 million in more than 10 cities such as Beijing and Shanghai, and all unmanned commercial operations were opened in Chongqing and Wuhan.

SEE ALSO: Baidu’s Robin Li: Next Commercial Stage of Autonomous Driving Is L4, Not L3

However, the challenges are great. The digital transformation of many fields in the real economy has not been completed, and digitalization itself has not brought about a significant improvement in efficiency. It will take time for intelligence to penetrate widely. Therefore, the commercialization of AI needs to be explored for some time.

On Baidu’s AI painting platform, an image is automatically generated by AI with “crisis” and “hope” as the keywords. There is no smooth sailing. For great enterprises and great creators, triumph comes after difficulties. Difficulties will stimulate innovation, and innovation is the real driving force of growth.