Our Privacy Statement & Cookie Policy

By continuing to browse our site you agree to our use of cookies, revised Privacy Policy and Terms of Use. You can change your cookie settings through your browser.

I agree

American AI plagiarism shows the importance of IP protection

Belunn Se

/CFP
/CFP

/CFP

Editor's note: Belunn Se, a special commentator on current affairs for CGTN, is a senior industry observer based in Shenzhen, China. The article reflects the author's opinions and not necessarily the views of CGTN.

Recently, a controversy involving AI model plagiarism has caused waves in the international tech community. A team from Stanford University released a large language model, Llama-3-V, which plagiarized the open-source work of Tsinghua University and Chinese tech firm Modelbest's MiniCPM-Llama3-V 2.5. This incident has sparked widespread concern in both academic circles and the public, highlighting issues of academic integrity, open-source community governance, and the international protection of China's innovative achievements in intellectual property (IP).

Modelbest released MiniCPM-Llama, its first edge-based multimodal large model, in 2021. In April 2024, it completed its Series A financing round of hundreds of millions of Chinese yuan, equivalent to tens of millions of U.S. dollars, and subsequently launched the MiniCPM-Llama3-V2.5, described as the world's most powerful edge-side multimodal model with eight billion parameters. This set new records in optical character recognition and image encoding speed.

This model is particularly well-suited for industrial application scenarios such as e-commerce, finance and education, unlike more generalized powerful models like ChatGPT, Claude and Google's Gemini, which are intended for broader general tasks like chatbots and text and photo processing.

On May 29, two Stanford University students and a student from the University of Southern California released their large language model Llama-3-V on a well-known open-source community. The model claimed to be able to train a multimodal model comparable to existing well-known models such as GPT-4V for just $500, and caused a sensation in the community.

However, it was discovered to be almost identical in architecture and code to MiniCPM-Llama3-V 2.5. On June 3, Li Dahai, CEO of Modelbest, released evidence confirming that the Stanford team's project used undisclosed Chinese ancient text data. He posted on his WeChat moments, saying that the two models showed a high degree of similarity in their responses and even shared the same errors, and that some relevant data had not yet been made public.

The two Stanford students then issued an apology on social media platform X on June 4, admitting to plagiarism and stating that they will remove the model.

The apology by two Stanford University students on X.
The apology by two Stanford University students on X.

The apology by two Stanford University students on X.

This incident has created a stir in both the academic community and society, serving as a stark reminder of the importance of academic integrity, which is the cornerstone of scholarly research. Upholding academic integrity is essential for promoting academic progress and benefiting society. However, the American team has seriously violated academic ethics, tarnishing their own reputation and negatively impacting the entire academic community.

The U.S. AI plagiarism also highlights the weakness of open-source framework governance, which is vital for the development and sharing of AI models. To prevent plagiarism and other unethical behaviors, a robust code review mechanism should be established.

Besides, it reflects the challenges and opportunities in protecting the IP of China's innovative achievements. Technological innovation is a crucial driving force for progress, and IP is its cornerstone. In the era of AI, intangible assets such as algorithms, models, and data hold significant value. If scientific research results are plagiarized, it will inevitably hinder the innovation process. Chinese companies have made remarkable progress in the AI sector and have developed substantial technological reserves that deserve proper and legal protection.

The scandal has sounded an alarm for Chinese AI companies to protect their IP on the international stage. China has made significant strides in AI innovation, but IP protection remains a critical issue that urgently needs to be addressed in international cooperation and competition, particularly against the backdrop of the fierce technological rivalry between China and the United States.

As pivotal forces in global technological development, both China and the United States possess immense potential and advantages in technological innovation and application. Strengthening cooperation between the two countries in science and technology in a fair and just environment will help promote global technological progress and achieve win-win outcomes. Only by respecting and protecting IP rights can long-term, healthy cooperation between both parties be ensured.

(If you want to contribute and have specific expertise, please contact us at opinions@cgtn.com. Follow @thouse_opinions on X, formerly Twitter, to discover the latest commentaries in the CGTN Opinion Section.)

Search Trends