Baidu CEO Robin Li Discusses the Future of AI Infrastructure and Innovation

Baidu CEO Robin Li emphasised the continued need for investment in data centres and cloud infrastructure to advance artificial intelligence (AI) models, despite the emergence of new AI systems that challenge the cost efficiency of large-scale models. Li pointed out that in order to create smarter AI models, there is a need for more computational power, which requires investment in hardware resources that are critical for training AI models, processing data, and generating predictions.

This comes as Chinese AI startup DeepSeek gains attention for developing language models that match the performance of leading systems like OpenAI’s GPT, yet use significantly less computing power. This has raised questions about whether the substantial spending on AI infrastructure is truly necessary.

Baidu, a pioneer in China’s AI landscape, launched its large language model Ernie in response to OpenAI’s ChatGPT release in 2022. However, despite claiming that Ernie is on par with GPT-4, the model has seen limited public adoption.

Li had previously predicted that China would not produce another company like OpenAI and had advocated for closed-source AI models as the primary path for development. However, at the summit, he acknowledged the unpredictable nature of innovation, particularly in light of DeepSeek's success.

Li also appeared to reconsider his stance on closed-source development, recognizing that open-source approaches could accelerate AI adoption. He noted that open-sourcing AI technology could spark widespread curiosity and experimentation, ultimately facilitating faster distribution and integration of the technology.