英語新聞丨錯(cuò)例都一樣！斯坦福學(xué)生團(tuán)隊(duì)致歉抄襲中國大模型

"'Fake it before you make it' is an ignoble product of Silicon Valley," said Christopher Manning, director of the Artificial Intelligence Laboratory at Stanford University, commenting on some researchers at the university who plagiarized the achievements by institutions such as China's Tsinghua University.

斯坦福大學(xué)人工智能實(shí)驗(yàn)室主任克里斯托弗·曼寧就該校某些研究人員抄襲中國清華大學(xué)等機(jī)構(gòu)成果的行為評論說:“‘作假，直至成功’，這是硅谷不光彩的文化?！?/p>

On May 29, a research team at Stanford University released a large model called Llama3-V, claiming it can achieve the same effects as large models such as GPT-4V with a pre-training cost of only US$500. The news was widely spread on social media and in the academic community of artificial intelligence.

5月29日，斯坦福大學(xué)一個(gè)研究團(tuán)隊(duì)發(fā)布名為Llama3-V的大模型，聲稱只要500美元的預(yù)訓(xùn)練成本，就能用它獲得比肩GPT-4V等著名大模型的效果。這一消息在社交媒體和人工智能學(xué)術(shù)界被廣泛傳播。

However, industry insiders soon suspected that the Standford team plagiarized the MiniCPM-Llama3-V 2.5 large model released by Tsinghua University and other Chinese institutions.Both Llama3-V and the MiniCPM-Llama3-V 2.5 large model are based on the open-source Llama3 large model. Still, the team in Tsinghua conducted unique training, including using the "Tsinghua Bamboo Slips," a collection of Chinese texts written on strips of bamboo which date back to the Warring States Period (475-221 BC), to train the model to recognize ancient Chinese characters.

但業(yè)內(nèi)人士很快懷疑，斯坦福團(tuán)隊(duì)抄襲了清華大學(xué)等中國機(jī)構(gòu)發(fā)布的MiniCPM-Llama3-V 2.5大模型。Llama3-V和MiniCPM-Llama3-V 2.5大模型都基于開源的Llama3大模型。不過清華團(tuán)隊(duì)進(jìn)行了獨(dú)特的訓(xùn)練，包括利用“清華簡”，這是一套寫在竹片上的中國文字，可以追溯到戰(zhàn)國時(shí)期（公元前475年至公元前221年），以訓(xùn)練模型識(shí)別古代漢字。

Tests show that the model released by the Stanford University team can also recognize the "Tsinghua Bamboo Slips."

測試顯示，斯坦福大學(xué)這個(gè)團(tuán)隊(duì)發(fā)布的大模型居然也能識(shí)別“清華簡”。

"We are quite sure that the Stanford team has plagiarized our big model research results," Liu Zhiyuan, a tenured associate professor of the Department of Computer Science at Tsinghua University, told Xinhua.

清華大學(xué)計(jì)算機(jī)系長聘副教授劉知遠(yuǎn)對新華社記者說:“已經(jīng)比較確信，斯坦福這個(gè)團(tuán)隊(duì)‘套殼’了我們的大模型研究成果?！?/p>

"The data we scanned and annotated word by word from the 'Tsinghua Bamboo Slips' has never been made public, and Llama3-V has shown the same ability to identify the 'Tsinghua Bamboo Slips,' even the error examples are the same," said Liu, who is also a member of the Tsinghua big model team.

劉知遠(yuǎn)說：“我們從‘清華簡’逐字掃描并標(biāo)注的數(shù)據(jù)集從未公開，而Llama3-V展現(xiàn)出了一模一樣的識(shí)別‘清華簡’能力，連做錯(cuò)的樣例都一樣。”劉知遠(yuǎn)是清華這個(gè)大模型團(tuán)隊(duì)成員。

As doubt accumulated, the Stanford team deleted the database and promotion articles on the Internet, Liu said, adding "from the evidence and their reactions, the nature of plagiarism has been relatively confirmed."

在質(zhì)疑聲發(fā)酵后，斯坦福大學(xué)團(tuán)隊(duì)刪除了網(wǎng)上發(fā)布的數(shù)據(jù)庫和宣傳文章，劉知遠(yuǎn)說，并補(bǔ)充“從證據(jù)和對方反應(yīng)來看，抄襲性質(zhì)已比較確定”。

Following Manning's criticism, two members of the Stanford team, Aksh Garg and Siddharth Sharma, formally apologized on social media.

在曼寧的批評之后，斯坦福團(tuán)隊(duì)的兩名成員阿克什·加格（Aksh Garg）和西達(dá)爾特·夏爾馬（Siddharth Sharma）在社交媒體上正式道歉。

"We've taken all references to Llama3-V down and we apologize once again for the inconvenience we may have caused," they said.

他們說:“我們已經(jīng)撤下了所有提及Llama3-V的內(nèi)容，我們再次為我們可能造成的不便表示歉意?！?/p>

Amid the current AI boom, this incident has aroused widespread attention. It shows that although the United States is leading in AI technologies overall, it is far from omnipotent.

在當(dāng)前人工智能熱潮中，這一事件引起了廣泛關(guān)注。這表明，盡管美國在人工智能技術(shù)方面總體領(lǐng)先，但它遠(yuǎn)非全能。

The Silicon Valley where Stanford University is located is considered to be the center of innovation in the United States. While having nurtured many advanced technologies, it has also cultivated a negative culture including the "fake it till you make it" ethos.

斯坦福大學(xué)所處的硅谷地區(qū)被認(rèn)為是美國科技創(chuàng)新的中心。既孕育了許多先進(jìn)技術(shù)，也培育出“作假，直至成功”等負(fù)面文化。

For example, Elizabeth Holmes, who dropped out of Stanford University to start a business, boasted that she had a disruptive technology that could draw finger blood to test diseases like cancers. She was regarded as a female Steve Jobs but was later found to have fooled everyone and was sentenced to imprisonment for fraud.

比如，從斯坦福大學(xué)退學(xué)創(chuàng)業(yè)的伊麗莎白·霍姆斯曾吹噓有顛覆性檢測技術(shù)可“抽指血查癌癥”。她一度被視為女史蒂夫·喬布斯，但后來發(fā)現(xiàn)她造假，并因欺詐被判入獄。

When Google's artificial intelligence model Gemini Pro was asked in Chinese who it was, it would answer that it was "Ernie Bot", a Chinese big model developed by Baidu. Industry insiders believe that the reason may be that Google "referenced" the relevant data of the large model "Ernie Bot" when training its large model.

當(dāng)谷歌的人工智能模型“雙子座”Pro版（Gemini Pro）被用中文問及是誰時(shí)，它會(huì)回答自己是百度開發(fā)的中文大模型“文心一言”。業(yè)界人士認(rèn)為，其原因可能是谷歌在訓(xùn)練大模型時(shí)“參考”了“文心一言”相關(guān)數(shù)據(jù)。

"China's AI research has an increasing influence," Liu said, noting the plagiarism incident reflects that "our innovative achievements are attracting international attention."

“中國人工智能研究的國際影響力越來越大”，劉知遠(yuǎn)說，并指出抄襲事件反映了“我們的創(chuàng)新成果正在受到國際關(guān)注”。

Overall, there is still a significant gap between China's research level and the world's top level, but in some specific segments such as AI innovation, China has rapidly grown into an important promoter, he added.

他補(bǔ)充說，總體而言，中國研究與國際頂尖成果仍有顯著差距，但在人工智能科技創(chuàng)新等一些具體領(lǐng)域，中國已快速成長為重要推動(dòng)者。

Plagiarism

剽竊

disruptive technology

顛覆性技術(shù)