Under the Banyan Tree - China and global AI rollout
Oct 26, 2023
auto_awesome
Frank Lee and Charlene Liu discuss China's AI development, including monetization strategies and the evolution of GPU chips. They explore revenue opportunities, regulation, and the differences between Chinese and international companies in the AI landscape.
The development of AI in China is progressing through three stages: investing in large language models, implementing inference capabilities, and developing edge AI.
Chinese AI companies face challenges in GPU capabilities and inventories, hindering the development and monetization of AI in China.
Deep dives
The Three Stages of AI Development
The first stage of AI development is focused on investing in large language models and training them using graphics processing chips (GPUs). This stage is currently dominated by NVIDIA as the sole provider of GPUs for AI services. The second stage, which is expected to start next year, will involve the implementation of inference capabilities, potentially using application-specific IC and ASIC chips. The third stage, referred to as the 'blue sky scenario,' involves the development of edge AI, where AI functionality can be performed on local devices, leading to potential growth in the smartphone and PC markets.
Monetization Strategies for Chinese AI Companies
Chinese AI companies adopt various monetization avenues, including AI computing services, model-as-a-service, and AI-generated content applications. AI computing services support AI applications, such as large language models training, by providing GPU usage to enhance processing abilities. Model-as-a-service enables the deployment of customized AI models, while AI-generated content applications focus on utilizing AI to enhance existing businesses and create new revenue opportunities. However, Chinese companies face challenges in GPU capabilities and inventories compared to their overseas counterparts.
Regulatory and Hardware Challenges for Chinese AI Industry
The Chinese AI industry is faced with regulatory changes, with China implementing a regulatory framework and approving language models, while the EU and the US are also addressing AI regulations. Chinese companies struggle with hardware limitations as they heavily depend on US chips for large-ignment models, while the US seeks to restrict chip access to China. This dependence on Western chips hampers the development and monetization of AI in China, creating a gap in hardware capabilities.