With a white shirt, a black suit and a thick head of hair, Robin Lee is particularly energetic among scientists and industry guests.
On November 15, Robin Lee, the founder, chairman and CEO of Baidu, as a mysterious guest, boarded the Shenzhen Xilihu Forum. However, the theme of his speech was not to "spark" the AI model, but to "pour cold water".
"The big model is a waste of social resources!" Robin Lee called himself "a practitioner of AI for more than ten years". When sharing the latest insights with the participating enterprises, he said that the opportunities of AI industry are in the original application.
This year's Xilihu Forum is themed on "Frontier, Transformation, and Connection", with AI big model as the core issue. However, Tencent, Huawei, and others did not appear in the venue. But Baidu went south to Shenzhen to preach.
Suppress first: Don't "roll up" the big model again and waste social resources
As the first technology company in China to release a big model, Robin Lee's first "cold" thought was to persuade people not to "roll" the big model.
"It is reported that as of October, 238 large models had been released in China, while in June, the number was 79, which is equivalent to a three fold increase in four months." Robin Lee first used a set of data on PPT to prove that China's large models are too "hot".
Based on this data, he began to pour cold water on enterprises that were already doing and preparing to do large models.
The big model itself is a technical foundation, similar to an operating system. Therefore, constantly developing various basic big models, I believe, is a great waste of social resources
Robin Lee also said that due to the lack of intelligence emerging ability, the value of special large models is actually very limited.
I see a phenomenon where many industries, enterprises, and even many cities are buying cards, hoarding chips, and establishing intelligent computing centers in order to train their own specialized large models from scratch. Little did they know that the large models refined in this way do not have the ability to emerge intelligently
However, he changed his tone and said, "Since its opening on August 31st, the API call volume of Wenxin Big Model has shown exponential growth. There are more than 200 big models in China that have made it to this list or that ranking, but the call volume is actually very small. The call volume of Wenxin Big Model is even more than the total call volume of these 200 big models
At the same time, Robin Lee also said that Baidu ERNIE Bot 4.0, the latest version launched last month, is no inferior to GPT-4 in terms of its comprehensive level, and believes that this version is "the most powerful so far".
Based on his own "most powerful" big model, Robin Lee said in his speech that "we don't need 100 big models".
Hou Yang: Encourage small and medium-sized enterprises to join the "Wenxin Ecology"
As a senior Internet tycoon, the logic of Robin Lee's preaching influences the industry. While declaring that "we don't need 100 big models", Robin Lee has a more important purpose for entrepreneurs and entrepreneurs in Shenzhen - to build ecology.
In this regard, Robin Lee revealed the purpose of his trip with a PPT: "In the AI native era, we need 1 million AI native applications."
In 'Cold Thinking', he launched 'Hot Drive'. Robin Lee began to actively encourage entrepreneurs and CEOs to work around Baidu. He believes that "embracing the AI era needs to be driven by CEOs and top leaders," and also states that "if CEOs do not actively lead this change, they are easily biased.
The implication is that SMEs should try to embrace AI native application platforms like ERNIE Bot. This may also indicate that the user group that the big model craves is the small and medium-sized enterprise owners.
At present, AI applications are already bringing changes to business operations, and a new form of startup company is emerging, characterized by being smaller and faster.
For example, Robin Lee said that for every 100 lines of code added by Baidu, which has tens of thousands of engineers, 20 lines are generated by AI, and the proportion is still growing rapidly. But the changes brought about by AI native applications are only just beginning.
Immediately, Robin Lee started a wave of "bringing goods": today, on the Baidu AI Cloud Qianfan model platform, more than 17000 enterprises have developed industrial models and solutions here. Not long ago, Baidu also launched a large model plugin platform. In the future, both individuals and businesses can quickly turn their data and capabilities into plugins through this platform. And this will greatly reduce the threshold for developers, which is conducive to building a prosperous AI native application ecosystem.
In the future, the way every enterprise deals with its own customers will be transformed into AI native applications, which will greatly enhance the competitiveness of the enterprise. Whether it is the improvement of enterprise competitiveness or the improvement of personal work efficiency, it is the driving force of economic growth
"The prosperity of AI application ecology will lead to economic prosperity." Robin Lee said that he hoped that everyone would take action to use it, understand it, experience it, and invest in the innovation of AI native applications, so as to jointly create an AI native era with a hundred flowers blooming and unlimited possibilities.
It is worth mentioning that Baidu has a lot of layout in AI scene applications in Shenzhen. In 2021, Pengcheng Laboratory and Baidu jointly released a knowledge enhancement model of one hundred billion yuan - "Pengcheng Baidu Wenxin"; In 2022, the Baidu Apollo autonomous driving travel service platform "Radish Run" will be launched in Nanshan District, providing citizens with autonomous driving demonstration application travel services; At the same time, Baidu has also trained a total of 45 teachers from 8 universities in Shenzhen, including Shenzhen University, Shenzhen University of Technology, Harbin Institute of Technology (Shenzhen), etc.