ByteDance said it terminated an intern in August for “maliciously interfering” with a model training project, following rife speculation on mainland Chinese social media platforms. Online reports about the resultant damage were exaggerated and inaccurate, it added.
The misconduct by the unnamed intern – who was part of the commercialisation technology team responsible for its advertising technology development – has been reported to industry associations and the relevant university for further actions, ByteDance said in a statement posted on Toutiao, its news aggregation platform, on Saturday.
The episode did not affect the company’s official commercial projects, online operations or its large language artificial intelligence (AI) models, according to the statement. The intern had no experience with the AI Lab and the person’s social media profile and some media reports contained inaccuracies, it added.
01:45
Chinese AI-generated cartoon series broadcast on state television
Chinese AI-generated cartoon series broadcast on state television
The clarification came after some social media posts alleged a ByteDance intern sabotaged the model training process with code because of dissatisfaction over resource allocation. The posts, which suggested the intern disrupted training involving a cluster of more than 8,000 H100 GPUs and caused tens of millions of dollars in losses, were overblown, ByteDance said.
05:03
How does China’s AI stack up against ChatGPT?
How does China’s AI stack up against ChatGPT?
Earlier this month, ByteDance launched Ola Friend, an open-ear wearable that allows users to interact with the company’s chatbot without using their smartphones.
An audio recording posted on an anonymous GitHub page called “JusticeFighter110 “on October 18 alleged the intern, an undergraduate from Beihang University and a master’s student at Peking University, admitted to uploading “malicious code” in the ByteDance project. A separate GitHub repository appeared on the same day, claiming the audio was faked.