At the X-Lake Forum in Shenzhen, Li, the co-founder, chair, and chief executive of Baidu, a prominent Chinese internet research company, disclosed that China had introduced 238 AI models by October, a significant increase from 79 in June. This growth was attributed to the popularity of LLMs, which are instrumental in training intelligent chatbots such as ChatGPT. However, Li noted a lack of widely recognized AI applications available to the general public.
Emphasizing the need for more AI-native programs utilizing existing models, Li stated, “China has many large designs but lacks sufficient AI-native applications.” He highlighted the importance of leveraging AI’s unparalleled capabilities, citing Tencent Holdings’ WeChat as an exemplary AI-native application.
Li criticized the excessive focus on developing foundation models, labeling it as a wasteful use of resources. He advocated for the creation of one million AI-native applications instead of numerous large models to drive innovation and progress in the field.
Robin Li Yanhong, the president and CEO of Baidu, is scheduled to speak at the monthly X-Lake Forum in Shenzhen on November 15, 2023.
LLMs, which are deep-learning AI algorithms trained on extensive datasets, play a crucial role in various tasks such as recognition, summarization, interpretation, and content generation.
To foster a thriving AI ecosystem and stimulate economic growth, Li underscored the importance of industry policies that promote the development of AI-native applications based on existing models.
Furthermore, Li cautioned against the trend of certain entities attempting to independently create artificial foundation models from scratch, citing the lack of feature-level data sets and coaching data as hindrances to the emergence of advanced AI capabilities.
In response to the increasing demand for AI chips in China, Intel has observed a notable uptick in market interest.
Li highlighted the significance of the number of parameters in LLMs, indicating their sophistication level. While models like OpenAI’s ChatGPT are trained on billions of characteristics, most Chinese LLMs available in the market have parameter ranges ranging from 6 to 13 billion.
He also suggested that government intervention could drive the adoption of large models in developing new AI applications.
Despite the absence of popular AI services like OpenAI’s ChatGPT in mainland China and Hong Kong, Chinese tech giants and IoT startups are actively competing to offer similar services. Notably, Microsoft Corp. has introduced its GPT-4 powered Bing Talk in the region.
The launch of GPT-4 Turbo by OpenAI is expected to intensify competition among international competitors.
Li highlighted the success of Baidu’s AI software, including the code-writing assistant Comate, emphasizing the exponential growth in Ernie Bot’s software programming software call volume since its inception in August.
In conclusion, Li expressed confidence in the future development of exceptional AI-native applications based on existing models, drawing parallels to the era that gave rise to popular apps like WeChat, Douyin, and Uber.
