[ad_1]
China’s open source AI models have been in the news recently for their strong performance in various AI tasks such as programming and “inference.”
However, they have also drawn criticism – Included From OpenAI employees – for censoring sensitive Chinese government topics, such as the Tiananmen Square massacre.
Clement Delange, CEO of HuggingFace, says he has similar concerns. Recently Podcast (in French), warned of unintended consequences for Western companies that rely on high-performance, open-source Chinese AI.
“If you create a chatbot and ask it a question about Tiananmen, it will not respond to you in the same way as if it were a system developed in France or the United States,” DeLange warned.
DeLange noted that if a country like China “becomes the most powerful in the field of artificial intelligence, it will be able to spread some aspects of culture that the Western world may not want to see spread.”
Delango It was mentioned previously Chinese AI is rapidly catching up with Western AI thanks to its embrace of the open source movement.
The strong concentration of top open source models coming from China is “a fairly new development that I’m a little worried about to be honest,” Delango warned on the podcast. “It is important that AI is distributed among all countries – so that no one or two countries are much stronger than others.”
HuggingFace is the world’s largest platform for AI modeling and a popular place for Chinese AI companies to showcase their latest MBA programs. In fact, the CTO of HuggingFace Announce This week, the default model on HuggingChat is Qwen2.5-72B-Instruct, developed by Chinese e-commerce giant Alibaba.
This model in particular does not appear to censor questions related to the Tiananmen Square massacre or others problems It is usually censored by the Chinese government.
A different model from Alibaba’s Qwen family available on HuggingChat, the QwQ-32B, explicitly does this when prompted by TC:
DeepSeek, another Chinese model that has caught on in the AI community for its reasoning capabilities, also imposes widespread censorship on topics deemed sensitive by the Chinese government, TechCrunch previously reported.
Chinese AI companies are in a difficult situation, as is the Chinese government Troops Their models “embody core socialist values” and are consistent with the already extensive system of censorship.
A HuggingFace spokesperson declined further comment but noted that Delangue recently anticipation China will start leading the global AI race in 2025.
[ad_2]