The US Stanford team copied China's open source model "Little steel cannon" publicly apologized for the unstoppable rise of China's AI big model

Recently, the Llama3-V open source model led by the Stanford University AI team was confirmed to copy the shell of China's Tsinghua University and the wall intelligence open source model "Little steel gun" MiniCPM-Llama3-V 2.5.

Two authors from Stanford's Llama3-V team, Sundhas Sharma and Aksha Garg, issued a formal apology to the Chinese MiniCPM team for the academic misconduct on a social platform.

Christopher David Manning, director of the Stanford Artificial Intelligence Lab, condemned the plagiarism in a post and praised MiniCPM, a Chinese open source model.

    The Stanford AI team advertised on the Web that it could train a large multimodal SOTA model beyond GPT-4V for as little as $500. However, the team's Llama3-V model uses a very similar model structure and code to the MiniCPM-Llama3-V2.5 recently released by China Wall Intelligence, with only some variable names modified. Llama3-V also has the same word segmentation as MiniCPM-Llama3-V 2.5, including the latter's newly defined special symbols.

The China Wall Intelligence team confirmed that the Stanford large model project Llama3-V, like MiniCPM, can identify the "Tsinghua Jane" Warring States ancient script, "not only the same, but also the same wrong." The Chinese research team spent months scanning and hand-marking the text verbatim from Tsinghua Scribes, which were never posted to confirm plagiarism.

Liu Zhiyuan, chief scientist of China Face Wall Intelligence and associate professor of Tsinghua University, said that the rapid development of artificial intelligence is inseparable from the open source sharing of global algorithms, data and models, so that people can always stand on the shoulders of SOTA to continue to move forward. The wall-facing open source MiniCPM-Llama3-V 2.5 uses the latest Llama3 as the language model base. The cornerstones of open source sharing are adherence to the open source protocol, trust in other contributors, and respect and homage to the work of those who came before, something that the Llama3-V team has certainly seriously undermined.

Since last year, China's big model has been open source one after another, transforming from a beneficiary to a contributor, and not shy of providing more excellent results of open source to the world. Up to giants such as Alibaba and Tencent, down to AI startups such as wall Intelligence, Zhipu AI, and Kunlun Tiangong, they are also activists in the open source community, contributing to the development of China's large model.

The incident is thought-provoking and has triggered a new round of global attention to the science and technology competition between China and the United States. Stanford University AI project team copied the open source model of China's large model company, from a point of view, this incident revealed the strength of China's AI research and development, more and more innovative products began to receive widespread international attention, China's development speed in large model research and development can be seen. China's innovation capabilities, application scenarios, algorithm optimization and data resources occupy an important position in the global competition for large models.

On the other hand, this incident also reveals a challenge in the field of AI, which is how to uphold and respect intellectual property rights. While sharing and borrowing are encouraged in the open source community, clear attribution and respect for the contributions of authors are basic ethical boundaries.

This incident has undoubtedly had a certain impact on the reputation of Stanford's AI team. In the field of research, academic integrity and respect for intellectual property rights are fundamental and important. The team copied others' open source models and tried to promote them as their own research results, which violated the principles of academic integrity and harmed the rights of the original authors.

Overall, for the global AI field, this is a thought-provoking event and an opportunity to learn and improve. It reveals China's strong strength in AI research and development, and also reminds people to respect and protect intellectual property rights while developing. In the future, we expect to see a global AI field that is fair, inclusive, and respectful of innovation.