Open Source has shown that when the obstacles to learning, using, sharing, and enhancing software systems are removed, everyone gains enormous advantages. Using licenses that follow the Open source Criteria yields several advantages.
In order for AI developers, deployers, and end users to benefit from the same advantages - autonomy, transparency, frictionless reuse, and collaborative improvement. A society must at least have the same fundamental freedoms as Open Source.
There is news circulating about Chinese open-sourcing AI models. Hugging Face CEO react to it and passes some statements.
Chinese Open Source AI Model
Recently, open source AI models from China have gained attention due to their impressive results on a variety of AI tasks, including reasoning and coding. They have also come under fire for banning subjects that are important to the Chinese government, such as the Tiananmen Square massacre, including from OpenAI staff.
Clement Delangue, CEO of Hugging Face, share these worries. He spoke against the unexpected effects of Western business constructing on top of successful Chinese AI in a recent podcast.
Delangue pointed out that if a nation like China dominates AI, it will have the ability to pass on cultural elements that the West would not want to see.
Hugging Face CEO Concerns
Delangue warned:
"A chatbot that you build and ask a question about Tiananmen will not react to you in the same manner as a system that was created in the United States or France."
The CEO had said before that because China has embraced the open source movement, its AI is rapidly catching up to Western AI.
The high concentration of popular open source models starting from China is a "fairly new development, and I'm a little worried about it, to be honest," Delangue remarked during the show. "It's crucial that AI be shared by all nations - that no one or two are significantly more powerful than the others."
Hugging Face and DeepSeek
Hugging Face is the largest AI model platform worldwide and a well-liked venue for Chinese AI firms to display their most recent LLMs. The CTO of Hugging Face actually revealed this week that Qwen2.5-72B-Instruct, created by the massive Chinese e-commerce company Alibaba, is the primary choice on HuggingChat.
Questions concerning the Tiananmen Square massacre and other topics that the Chinese government usually censors do not seem to be restricted by this specific model.
According to a recent TechCrunch article, DeepSeek, another Chinese model that gained popularity in the AI industry for its reasoning skills, also heavily filters content that the Chinese government finds cautious.
The Chinese government requires AI models to "represent basic socialism values" and submit to its existing strict censorship system, which puts Chinese AI companies in a difficult position.
A Hugging Face representative noted that Delangue recently estimated that China would begin to dominate the global AI competition in 2025 but refused to talk further.