On the morning of November 15th, the 2023 Xilihu Forum opened at the International Conference Center of Shenzhen University City, Guangdong. Multiple business leaders and educational elites gathered at the Xilihu Forum to share their observations and reflections on cutting-edge technology.
At the opening ceremony, Robin Lee, the founder, chairman and CEO of Baidu, as the "mysterious guest" announced the day before the opening ceremony, launched a speech themed "AI Original Age:" Cold "Thinking and" Hot "Driving" around the hottest topic of AI big model.
Robin Lee believes that there are many big models in China, but there are very few AI native applications developed based on the big models. In his view, the hallmark of humanity entering the AI era is not the emergence of many large models, but the emergence of many AI native applications. He used the era of personal computers and the era of mobile internet as examples to explain: in the era of personal computers, there was basically only one operating system, Windows, but there were many software developed based on Windows; In the era of mobile internet, the mainstream operating systems are only Android and iOS, while there are as many as 8 million mobile applications. The same is true in the era of big models. Big models themselves are a foundation, similar to operating systems, and ultimately developers rely on a few large models to develop various native applications.
"Repeated development of large basic models is a great waste of social resources. In the AI native era, we need 1 million AI native applications, but we do not need 100 large models." Robin Lee said.
Robin Lee pointed out that due to the lack of intelligence emerging ability, the value of special large models is very limited. Because only when the parameter scale of the model is large enough, the amount of training data is sufficient, and continuous investment and iteration are carried out, can intelligent emergence be generated, and large models can have the ability to bypass by analogy. In his view, the industrialization mode of large models should combine the general capabilities of basic models with professional knowledge in the industry field, that is, large models are combined with small models, specialized small models have fast response, low cost, and large models are more intelligent, which can be used as a backup.
In the speech, Robin Lee also focused on sharing the three "hot" drivers of AI's original era. First of all, the powerful basic model will drive the explosion of AI native applications. He believes that no AI native applications that can benchmark WeChat, Tiktok, and Uber in the mobile era have been produced in China or the United States, and there must be excellent AI native applications developed based on the big model in the AI native era. Secondly, embracing the AI era requires the CEO and top executives to drive it, as only CEOs are concerned about whether new technologies have a positive impact on key indicators of their business. Finally, a prosperous AI native application ecosystem will drive economic growth. The AI industry is demand driven, and efforts should be made on the demand side and application layer to encourage enterprises to use big models to develop AI native applications and promote industrial development through market means.
It is understood that Baidu first released ERNIE Bot product based on Wenxin big model 3.0 on March 16 this year, and then quickly iterated, and released Wenxin 4.0 version last month. Robin Lee introduced in his speech that 4.0 is the most powerful cultural model so far, and its ability in understanding, generation, logic and memory has been significantly improved. At the same time, Baidu has also undergone AI native refactoring of its various product lines such as Baidu Search and Baidu Wenku.
"I hope that everyone will take action, invest in the innovation of AI native applications, and jointly create an AI native era with a hundred flowers blooming and unlimited possibilities," said Robin Lee.