On Wednesday, Hong Kong (HK) cautioned companies to prioritize the development of practical applications due to the rapid advancement of large language models in China. Robin Li emphasized this point during an industry forum in Shenzhen, noting the lack of viable business models among companies engaged in building extensive vocabulary models, amidst concerns of a potential market downturn.
The emergence of generative artificial intelligence (AI) has captured significant attention in China following the introduction of the ChatGPT chatbot by OpenAI late last year. This development has spurred both established enterprises and startups to explore opportunities in the sector. Li observed a trend in China where numerous entities across various sectors are investing in hardware, chips, and computing infrastructure to train custom large models from the ground up.
While emphasizing the importance of big language models as foundational elements akin to operating systems, Li highlighted the necessity for developers to leverage existing models to create diverse applications efficiently. He criticized the continuous redevelopment of core models as a significant drain on societal resources.
Despite the proliferation of large models in China, Li pointed out the limited number of AI applications utilizing these resources. A third-party report referenced by Li indicated a substantial increase in the release of big language models, with 238 versions introduced in October compared to just 79 in June.
Baidu’s proprietary large speech model, Ernie, obtained official approval for public release in August, joining a cohort of innovative products in the market. The most recent variant, Ernie 4.0, an advanced conceptual AI model by Baidu, was unveiled last month, following its initial launch in March.
