By Diego Ramos / Original to ScheerPost
Two Stanford University undergraduates apologized in a post on X following claims that their newly launched AI model, Llama 3-V, had a structure and code that was almost identical to an open-source Chinese model.
Following a popular launch last week, Llama 3-V faced heavy scrutiny for its striking similarities to MiniCPM-Llama3-V 2.5, an AI model developed by Tsinghua University’s Natural Language Processing Lab and the Beijing-based AI start-up ModelBest.
Stanford computer science undergraduates Aksh Garg and Siddharth Sharma issued a apology on X to the authors of MiniCPM, acknowledging in a since-deleted Medium post, “We realized that our architecture is very similar.”
Garg and Sharma also claimed another researcher, Mustafa Aljadery, who is not from Stanford, wrote the code for Llama 3-V but lost contact with him prior to issuing the apology.
Llama 3-V was intended to rival popular AI models such as GPT4-V, Gemini Ultra and Claude Opus and even made it onto Hugging Face’s, a popular artificial intelligence platform, top five trending list.
According to a South China Morning Post report, Liu Zhiyuan, co-founder of ModelBest, explained, “he was ‘relatively sure’ that the new model had stolen from their project.”
Liu went on to explain that the Chinese MiniCPM model had access to a unique dataset—bamboo slips from the Warring States Period—which distinguished its recognition abilities from other softwares. Liu noted that this same recognition ability was evident in Llama 3-V.
Liu also criticized Garg and Sharma, expressing that “the cornerstones of open-source sharing were adhering to protocols, trusting other contributors, and respecting and acknowledging the work of pioneers, which the Stanford team had ‘seriously undermined.’”
Please share this story and help us grow our network!
Diego Ramos
Diego Ramos, ScheerPost Special Projects Editor and New York bureau chief, is a journalist from Queens, NY. He graduated from the University of Southern California in 2022 with a bachelor’s degree in journalism. He has previously worked at BuzzFeed News and was managing editor of Annenberg News at USC. He’s covered and researched myriad topics including war, politics, psychedelic research and sports.
Editor’s Note: At a moment when the once vaunted model of responsible journalism is overwhelmingly the play thing of self-serving billionaires and their corporate scribes, alternatives of integrity are desperately needed, and ScheerPost is one of them. Please support our independent journalism by contributing to our online donation platform, Network for Good, or send a check to our new PO Box. We can’t thank you enough, and promise to keep bringing you this kind of vital news.
You can also make a donation to our PayPal or subscribe to our Patreon.