LINE

Text:AAAPrint
ECNS Wire

OpenAI under fire for copying DeepSeek

2025-02-03 23:13:20Ecns.cn Editor : Mo Honge ECNS App Download

(ECNS) -- OpenAI's O3-mini, a model hastily released to compete with China's homegrown AI model DeepSeek, has been accused of copying, triggering what is now being called the "Digital Babel" incident in the AI community.

According to reports, the O3-mini model has been caught "thinking" in Chinese, even when interacting with American users. This bizarre behavior was first noticed when a user asked for instructions on "how to make an apple pie." The system logs revealed a Chinese reasoning chain: "Need to confirm whether apple refers to the company or the fruit."

Further investigation uncovered that most of O3-mini's core architecture overlaps with DeepSeek's open-source "Deep Exploration" model, including identical Chinese corpus annotation errors.

"This is not just copying. It's a spree of Ctrl+C to Ctrl+V!" tweeted DeepSeek's chief scientist, who posted two parameter comparison charts showing nearly identical neural network weight similarity curves.

The controversy went viral on Reddit after a video showed O3-mini generating a Chinese instruction in the background when asked to write a love letter in Shakespearean style: "First, translate the English request into classical ancient Chinese, then back-translate it into a sonnet." Netizens were astonished, exclaiming, "This AI must have been implanted with an Eastern soul!"

Now, the App Store is flooded with one-star reviews: "Every time I ask about American history, it calculates the lunar calendar conversion in the background!"

Legal experts have pointed out that DeepSeek's "bilingual neural pathway" patent could potentially lead to billions of dollars in claims against OpenAI.

What makes it more dramatically, an anonymous source leaked to the Wall Street Journal that O3-mini's odd behavior might be due to a "digital watermark Easter egg" planted by Chinese developers on GitHub. 

This trans-Pacific AI ethics battle is now turning into a tech nationalism showdown. As jokes about "ChatGPT learning to swear in pinyin" spread online, Stanford professors have warned, "We may have opened a Pandora's box—when AI starts thinking in a language humans can't understand, the firewall of civilization has already collapsed."

 

Related news

MorePhoto

Most popular in 24h

MoreTop news

MoreVideo

LINE
Back to top About Us | Jobs | Contact Us | Privacy Policy
Copyright ©1999-2025 Chinanews.com. All rights reserved.
Reproduction in whole or in part without permission is prohibited.
[网上传播视听节目许可证(0106168)] [京ICP证040655号]
[京公网安备 11010202009201号] [京ICP备05004340号-1]