Debate in China between open-source and closed-source AI models, with key players promoting open-source.
From Caixin Global: 2024-06-25 10:55:03
The debate over large language models (LLMs) in China heats up as CEO Zhou Hongyi criticizes claims that open-source LLMs are inferior to closed-source ones, indirectly responding to Baidu CEO, Robin Li. Baidu offers a local alternative to OpenAI’s ChatGPT, while other companies like Alibaba and Beijing Academy of AI support open-source LLMs.
Real-world success stories emerge with designers and developers like Liu Ming and Professor Xue Dong using open-source LLMs for design tasks and developing chatbots, respectively. Alibaba’s open-source model, Qwen-72B, showcases the accuracy and affordability of open-source solutions over closed-source alternatives.
Alibaba’s ModelScope platform, hosting over 4,000 deployable AI models, engages millions of developers globally. Alibaba’s CTO emphasizes an open ecosystem approach for LLMs, contrary to Baidu’s strategy. Despite promoting cloud services, Alibaba focuses on open-sourcing LLMs for wider accessibility and innovation.
Baidu CEO Li argues that transforming open-source LLMs into profitable ventures requires significant investment. Baidu introduces AI development toolkits to facilitate building applications using their closed-source LLM ERNIE. Concerns over data ownership and accessibility to Baidu’s APIs arise, prompting the need for comprehensive development toolkits and computing resources.
Market observers highlight the importance of blending computing resources and development toolkits for effective AI applications. Former GitLab China employee, Chen Ran, identifies a business opportunity in this space, launching OpenCSG to offer tools for LLM development and deployment, including a computing resources trading platform expected to generate revenue.
The ongoing discussion between open-source and closed-source models evolves with the availability of resources and development tools catering to the needs of AI developers. The decision on which model to adopt ultimately hinges on factors such as cost, accuracy, flexibility, and ease of use in different applications.
Read more at Caixin Global: In Depth: China Debates Pros and Cons of Open-Source AI Models